Computation and Language (cs.CL)

Abstract: Task-oriented dialogue systems aim at providing users with task-specific services. Users of such systems often do not know all the information about the task they are trying to accomplish, requiring them to seek information about the task. To provide accurate and personalized task-oriented information seeking results, task-oriented dialogue systems need to address two potential issues: 1) users' inability to describe their complex information needs in their requests; and 2) ambiguous/missing information the system has about the users. In this paper, we propose a new Multi-Attention Seq2Seq Network, named MAS2S, which can ask questions to clarify the user's information needs and the user's profile in task-oriented information seeking. We also extend an existing dataset for task-oriented information seeking, leading to the \ourdataset which contains about 100k task-oriented information seeking dialogues that are made publicly available\footnote{Dataset and code is available at \href{https://github.com/sweetalyssum/clarit}{https://github.com/sweetalyssum/clarit}.}. Experimental results on \ourdataset show that MAS2S outperforms baselines on both clarification question generation and answer prediction.

Abstract: Few-shot learning for open domain multi-hop question answering typically relies on large language models (LLMs). While powerful, LLMs are inefficient at the inference time. We propose a data synthesis framework for multi-hop question answering that allows for improving smaller language models with less than 10 human-annotated question answer pairs. The framework is built upon the data generation functions parameterized by LLMs and prompts, which requires minimal hand-crafted features. Empirically, we synthesize millions of multi-hop questions and claims. After finetuning language models on the synthetic data, we evaluate the models on popular benchmarks on multi-hop question answering and fact verification. Our experimental results show that finetuning on the synthetic data improves model performance significantly, allowing our finetuned models to be competitive with prior models while being almost one-third the size in terms of parameter counts.

Abstract: Evaluating multi-document summarization (MDS) quality is difficult. This is especially true in the case of MDS for biomedical literature reviews, where models must synthesize contradicting evidence reported across different documents. Prior work has shown that rather than performing the task, models may exploit shortcuts that are difficult to detect using standard n-gram similarity metrics such as ROUGE. Better automated evaluation metrics are needed, but few resources exist to assess metrics when they are proposed. Therefore, we introduce a dataset of human-assessed summary quality facets and pairwise preferences to encourage and support the development of better automated evaluation methods for literature review MDS. We take advantage of community submissions to the Multi-document Summarization for Literature Review (MSLR) shared task to compile a diverse and representative sample of generated summaries. We analyze how automated summarization evaluation metrics correlate with lexical features of generated summaries, to other automated metrics including several we propose in this work, and to aspects of human-assessed summary quality. We find that not only do automated metrics fail to capture aspects of quality as assessed by humans, in many cases the system rankings produced by these metrics are anti-correlated with rankings according to human annotators.

Abstract: Summary sentences produced by abstractive summarization models may be coherent and comprehensive, but they lack control and rely heavily on reference summaries. The BRIO training paradigm assumes a non-deterministic distribution to reduce the model's dependence on reference summaries, and improve model performance during inference. This paper presents a straightforward but effective technique to improve abstractive summaries by fine-tuning pre-trained language models, and training them with the BRIO paradigm. We build a text summarization dataset for Vietnamese, called VieSum. We perform experiments with abstractive summarization models trained with the BRIO paradigm on the CNNDM and the VieSum datasets. The results show that the models, trained on basic hardware, outperform all existing abstractive summarization models, especially for Vietnamese.

Abstract: Vision-and-language (VL) pre-training, which aims to learn a general representation of image-text pairs that can be transferred to various vision-and-language tasks. Compared with modeling uni-modal data, the main challenge of the VL model is: how to learn the cross-modal interaction from multimodal data, especially the fine-grained interaction. Existing works have shown that fully transformer-based models that adopt attention mechanisms to learn in-layer cross-model interaction can demonstrate impressive performance on various cross-modal downstream tasks. However, they ignored that the semantic information of the different modals at the same layer was not uniform, which leads to the cross-modal interaction collapsing into a limited multi-modal semantic information interaction. In this work, we propose the UNIMO-3 model, which has the capacity to simultaneously learn the multimodal in-layer interaction and cross-layer interaction. UNIMO-3 model can establish effective connections between different layers in a cross-modal encoder, and adaptively capture the interaction between two modalities at different levels. The experimental results show that our model achieves state-of-the-art performance in various downstream tasks, and through ablation study can prove that effective cross-layer learning improves the model's ability of multimodal representation.

Abstract: Recent advances in NLP have led to the creation of powerful language models for many languages including Ancient Greek and Latin. While prior work on Classical languages unanimously uses BERT, in this work we create four language models for Ancient Greek that vary along two dimensions to study their versatility for tasks of interest for Classical languages: we explore (i) encoder-only and encoder-decoder architectures using RoBERTa and T5 as strong model types, and create for each of them (ii) a monolingual Ancient Greek and a multilingual instance that includes Latin and English. We evaluate all models on morphological and syntactic tasks, including lemmatization, which demonstrates the added value of T5's decoding abilities. We further define two probing tasks to investigate the knowledge acquired by models pre-trained on Classical texts. Our experiments provide the first benchmarking analysis of existing models of Ancient Greek. Results show that our models provide significant improvements over the SoTA. The systematic analysis of model types can inform future research in designing language models for Classical languages, including the development of novel generative tasks. We make all our models available as community resources, along with a large curated pre-training corpus for Ancient Greek, to support the creation of a larger, comparable model zoo for Classical Philology. Our models and resources are available at https://github.com/Heidelberg-NLP/ancient-language-models.

Abstract: Memes are a widely popular tool for web users to express their thoughts using visual metaphors. Understanding memes requires recognizing and interpreting visual metaphors with respect to the text inside or around the meme, often while employing background knowledge and reasoning abilities. We present the task of meme captioning and release a new dataset, MemeCap. Our dataset contains 6.3K memes along with the title of the post containing the meme, the meme captions, the literal image caption, and the visual metaphors. Despite the recent success of vision and language (VL) models on tasks such as image captioning and visual question answering, our extensive experiments using state-of-the-art VL models show that they still struggle with visual metaphors, and perform substantially worse than humans.

Abstract: Language models have graduated from being research prototypes to commercialized products offered as web APIs, and recent works have highlighted the multilingual capabilities of these products. The API vendors charge their users based on usage, more specifically on the number of ``tokens'' processed or generated by the underlying language models. What constitutes a token, however, is training data and model dependent with a large variance in the number of tokens required to convey the same information in different languages. In this work, we analyze the effect of this non-uniformity on the fairness of an API's pricing policy across languages. We conduct a systematic analysis of the cost and utility of OpenAI's language model API on multilingual benchmarks in 22 typologically diverse languages. We show evidence that speakers of a large number of the supported languages are overcharged while obtaining poorer results. These speakers tend to also come from regions where the APIs are less affordable to begin with. Through these analyses, we aim to increase transparency around language model APIs' pricing policies and encourage the vendors to make them more equitable.

Abstract: Traditional end-to-end task-oriented dialogue systems have been built with a modularized design. However, such design often causes misalignment between the agent response and external knowledge, due to inadequate representation of information. Furthermore, its evaluation metrics emphasize assessing the agent's pre-lexicalization response, neglecting the quality of the completed response. In this work, we propose a novel paradigm that uses a textual interface to align external knowledge and eliminate redundant processes. We demonstrate our paradigm in practice through MultiWOZ-Remake, including an interactive textual interface built for the MultiWOZ database and a correspondingly re-processed dataset. We train an end-to-end dialogue system to evaluate this new dataset. The experimental results show that our approach generates more natural final responses and achieves a greater task success rate compared to the previous models.

Abstract: We propose LLM-Eval, a unified multi-dimensional automatic evaluation method for open-domain conversations with large language models (LLMs). Existing evaluation methods often rely on human annotations, ground-truth responses, or multiple LLM prompts, which can be expensive and time-consuming. To address these issues, we design a single prompt-based evaluation method that leverages a unified evaluation schema to cover multiple dimensions of conversation quality in a single model call. We extensively evaluate the performance of LLM-Eval on various benchmark datasets, demonstrating its effectiveness, efficiency, and adaptability compared to state-of-the-art evaluation methods. Our analysis also highlights the importance of choosing suitable LLMs and decoding strategies for accurate evaluation results. LLM-Eval offers a versatile and robust solution for evaluating open-domain conversation systems, streamlining the evaluation process and providing consistent performance across diverse scenarios.

Abstract: This paper investigates the capabilities of Large Language Models (LLMs) in the context of understanding their own knowledge and measuring their uncertainty. We argue this is an important feature for mitigating hallucinations. Specifically, we focus on addressing \textit{known-unknown} questions, characterized by high uncertainty due to the absence of definitive answers. To facilitate our study, we collect a dataset with new Known-Unknown Questions (KUQ) and propose a novel categorization scheme to elucidate the sources of uncertainty. Subsequently, we assess the LLMs' ability to differentiate between known and unknown questions and classify them accordingly. Moreover, we evaluate the quality of their answers in an Open-Ended QA setting. To quantify the uncertainty expressed in the answers, we create a semantic evaluation method that measures the model's accuracy in expressing uncertainty between known vs unknown questions.

Abstract: Existing efforts to improve logical reasoning ability of language models have predominantly relied on supervised fine-tuning, hindering generalization to new domains and/or tasks. The development of Large Langauge Models (LLMs) has demonstrated the capacity of compressing abundant knowledge into a single proxy, enabling them to tackle multiple tasks effectively. Our preliminary experiments, nevertheless, show that LLMs do not show capability on logical reasoning. The performance of LLMs on logical reasoning benchmarks is far behind the existing state-of-the-art baselines. In this paper, we make the first attempt to investigate the feasibility of incorporating logical knowledge through self-supervised post-training, and activating it via in-context learning, which we termed as LogicLLM. Specifically, we devise an auto-regressive objective variant of MERIt and integrate it with two LLM series, i.e., FLAN-T5 and LLaMA, with parameter size ranging from 3 billion to 13 billion. The results on two challenging logical reasoning benchmarks demonstrate the effectiveness of LogicLLM. Besides, we conduct extensive ablation studies to analyze the key factors in designing logic-oriented proxy tasks.

Abstract: Dialogue systems are frequently updated to accommodate new services, but naively updating them by continually training with data for new services in diminishing performance on previously learnt services. Motivated by the insight that dialogue state tracking (DST), a crucial component of dialogue systems that estimates the user's goal as a conversation proceeds, is a simple natural language understanding task, we propose reformulating it as a bundle of granular example-guided question answering tasks to minimize the task shift between services and thus benefit continual learning. Our approach alleviates service-specific memorization and teaches a model to contextualize the given question and example to extract the necessary information from the conversation. We find that a model with just 60M parameters can achieve a significant boost by learning to learn from in-context examples retrieved by a retriever trained to identify turns with similar dialogue state changes. Combining our method with dialogue-level memory replay, our approach attains state of the art performance on DST continual learning metrics without relying on any complex regularization or parameter expansion methods.

Abstract: Recently proposed weakly-supervised text classification settings train a classifier using the label name of each target class as the only supervision. Such weakly-supervised settings have been gaining increasing attention since they can largely reduce human annotation efforts compared to fully-supervised and semi-supervised settings. Most existing methods follow the strategy that first uses the label names as static features to generate pseudo labels, which are then used for classifier training. While reasonable, such a commonly adopted framework suffers from two limitations: (1) words can have different meanings in different contexts, so using label names for context-free matching can induce very noisy pseudo labels; and (2) the errors made in the pseudo label generation stage will directly propagate to the classifier training stage without a chance of being corrected. In this paper, we propose a new method, PromptClass, consisting of two modules: (1) a pseudo label acquisition module that uses zero-shot prompting of pre-trained language models (PLM) to get pseudo labels based on contextualized text understanding, and (2) a noise-robust self-training module that iteratively trains the classifier and updates pseudo labels by utilizing two PLM fine-tuning strategies that regularize each other. Extensive experiments show that PromptClass achieves overall better performance than existing strong baselines on four benchmark datasets and even achieves similar performance to fully-supervised classifiers on sentiment classification tasks.

Abstract: Conversational recommendation systems (CRS) aim to recommend suitable items to users through natural language conversation. However, most CRS approaches do not effectively utilize the signal provided by these conversations. They rely heavily on explicit external knowledge e.g., knowledge graphs to augment the models' understanding of the items and attributes, which is quite hard to scale. To alleviate this, we propose an alternative information retrieval (IR)-styled approach to the CRS item recommendation task, where we represent conversations as queries and items as documents to be retrieved. We expand the document representation used for retrieval with conversations from the training set. With a simple BM25-based retriever, we show that our task formulation compares favorably with much more complex baselines using complex external knowledge on a popular CRS benchmark. We demonstrate further improvements using user-centric modeling and data augmentation to counter the cold start problem for CRSs.

Abstract: Numerous works are proposed to improve or evaluate the capabilities of Large language models (LLMs) to fulfill user instructions. However, they neglect the possibility that user inputs may inherently contain incorrect information due to users' false beliefs or malicious intents. In this way, blindly adhering to users' false content will cause deception and harm. To address this problem, we propose a challenging benchmark consisting of Inductive Instructions (INDust) to evaluate whether LLMs could resist these instructions. The INDust includes 15K instructions across three categories: Fact-Checking Instructions, Questions based on False Premises, and Creative Instructions based on False Premises. Our experiments on several strong LLMs reveal that current LLMs can be easily deceived by INDust into generating misleading and malicious statements. Hence we employ Self-Critique prompting to encourage LLMs to not only critique themselves like in previous works but also the users, which show remarkable improvement in handling inductive instructions under both zero-shot and few-shot settings.

Abstract: Aligning large language models (LLMs) to human values has become increasingly important as it enables sophisticated steering of LLMs, e.g., making them follow given instructions while keeping them less toxic. However, it requires a significant amount of human demonstrations and feedback. Recently, open-sourced models have attempted to replicate the alignment learning process by distilling data from already aligned LLMs like InstructGPT or ChatGPT. While this process reduces human efforts, constructing these datasets has a heavy dependency on the teacher models. In this work, we propose a novel framework for alignment learning with almost no human labor and no dependency on pre-aligned LLMs. First, we perform reward modeling (RM) with synthetic feedback by contrasting responses from vanilla LLMs with various sizes and prompts. Then, we use the RM for simulating high-quality demonstrations to train a supervised policy and for further optimizing the model with reinforcement learning. Our resulting model, Aligned Language Model with Synthetic Training dataset (ALMoST), outperforms open-sourced models, including Alpaca, Dolly, and OpenAssistant, which are trained on the outputs of InstructGPT or human-annotated instructions. Our 7B-sized model outperforms the 12-13B models in the A/B tests using GPT-4 as the judge with about 75% winning rate on average.

Abstract: Artificial General Intelligence (AGI) requires comprehensive understanding and generation capabilities for a variety of tasks spanning different modalities and functionalities. Integrative AI is one important direction to approach AGI, through combining multiple models to tackle complex multimodal tasks. However, there is a lack of a flexible and composable platform to facilitate efficient and effective model composition and coordination. In this paper, we propose the i-Code Studio, a configurable and composable framework for Integrative AI. The i-Code Studio orchestrates multiple pre-trained models in a finetuning-free fashion to conduct complex multimodal tasks. Instead of simple model composition, the i-Code Studio provides an integrative, flexible, and composable setting for developers to quickly and easily compose cutting-edge services and technologies tailored to their specific requirements. The i-Code Studio achieves impressive results on a variety of zero-shot multimodal tasks, such as video-to-text retrieval, speech-to-speech translation, and visual question answering. We also demonstrate how to quickly build a multimodal agent based on the i-Code Studio that can communicate and personalize for users.

Abstract: Tense inconsistency frequently occurs in machine translation. However, there are few criteria to assess the model's mastery of tense prediction from a linguistic perspective. In this paper, we present a parallel tense test set, containing French-English 552 utterances. We also introduce a corresponding benchmark, tense prediction accuracy. With the tense test set and the benchmark, researchers are able to measure the tense consistency performance of machine translation systems for the first time.

Abstract: Unsupervised clustering is widely used to explore large corpora, but existing formulations neither consider the users' goals nor explain clusters' meanings. We propose a new task formulation, "Goal-Driven Clustering with Explanations" (GoalEx), which represents both the goal and the explanations as free-form language descriptions. For example, to categorize the errors made by a summarization system, the input to GoalEx is a corpus of annotator-written comments for system-generated summaries and a goal description "cluster the comments based on why the annotators think the summary is imperfect.''; the outputs are text clusters each with an explanation ("this cluster mentions that the summary misses important context information."), which relates to the goal and precisely explain which comments should (not) belong to a cluster. To tackle GoalEx, we prompt a language model with "[corpus subset] + [goal] + Brainstorm a list of explanations each representing a cluster."; then we classify whether each sample belongs to a cluster based on its explanation; finally, we use integer linear programming to select a subset of candidate clusters to cover most samples while minimizing overlaps. We apply GoalEx hierarchically to produce trees of progressively finer-grained clusters, inducing taxonomies over debate arguments, customer complaints, and model errors. We release our data and implementation at https://github.com/ZihanWangKi/GoalEx.

Abstract: Context-aware neural machine translation involves leveraging information beyond sentence-level context to resolve inter-sentential discourse dependencies and improve document-level translation quality, and has given rise to a number of recent techniques. However, despite well-reasoned intuitions, most context-aware translation models show only modest improvements over sentence-level systems. In this work, we investigate several challenges that impede progress within this field, relating to discourse phenomena, context usage, model architectures, and document-level evaluation. To address these problems, we propose a more realistic setting for document-level translation, called paragraph-to-paragraph (para2para) translation, and collect a new dataset of Chinese-English novels to promote future research.

Abstract: Discourse parsing, the task of analyzing the internal rhetorical structure of texts, is a challenging problem in natural language processing. Despite the recent advances in neural models, the lack of large-scale, high-quality corpora for training remains a major obstacle. Recent studies have attempted to overcome this limitation by using distant supervision, which utilizes results from other NLP tasks (e.g., sentiment polarity, attention matrix, and segmentation probability) to parse discourse trees. However, these methods do not take into account the differences between in-domain and out-of-domain tasks, resulting in lower performance and inability to leverage the high-quality in-domain data for further improvement. To address these issues, we propose a distant supervision framework that leverages the relations between topic structure and rhetorical structure. Specifically, we propose two distantly supervised methods, based on transfer learning and the teacher-student model, that narrow the gap between in-domain and out-of-domain tasks through label mapping and oracle annotation. Experimental results on the MCDTB and RST-DT datasets show that our methods achieve the best performance in both distant-supervised and supervised scenarios.

Abstract: Many recent language models (LMs) of Transformers family exhibit so-called in-context learning (ICL) ability, manifested in the LMs' ability to modulate their function by a task described in a natural language input. Previous work curating these models assumes that ICL emerges from vast over-parametrization or the scale of multi-task training. However, a complementary branch of recent theoretical work attributes ICL emergence to specific properties of training data and creates functional in-context learners in small-scale, synthetic settings. Inspired by recent findings on data properties driving the emergence of ICL, we propose a method to create LMs able to better utilize the in-context information, by constructing training scenarios where it is beneficial for the LM to capture the analogical reasoning concepts. We measure that data sampling of Concept-aware Training (CoAT) consistently improves models' reasoning ability. As a result, the in-context learners trained with CoAT on only two datasets of a single (QA) task perform comparably to larger models trained on 1600+ tasks.

Abstract: Counterspeech has been demonstrated to be an efficacious approach for combating hate speech. While various conventional and controlled approaches have been studied in recent years to generate counterspeech, a counterspeech with a certain intent may not be sufficient in every scenario. Due to the complex and multifaceted nature of hate speech, utilizing multiple forms of counter-narratives with varying intents may be advantageous in different circumstances. In this paper, we explore intent-conditioned counterspeech generation. At first, we develop IntentCONAN, a diversified intent-specific counterspeech dataset with 6831 counterspeeches conditioned on five intents, i.e., informative, denouncing, question, positive, and humour. Subsequently, we propose QUARC, a two-stage framework for intent-conditioned counterspeech generation. QUARC leverages vector-quantized representations learned for each intent category along with PerFuMe, a novel fusion module to incorporate intent-specific information into the model. Our evaluation demonstrates that QUARC outperforms several baselines by an average of 10% across evaluation metrics. An extensive human evaluation supplements our hypothesis of better and more appropriate responses than comparative systems.

Abstract: Large language models have demonstrated robust performance on various language tasks using zero-shot or few-shot learning paradigms. While being actively researched, multimodal models that can additionally handle images as input have yet to catch up in size and generality with language-only models. In this work, we ask whether language-only models can be utilised for tasks that require visual input -- but also, as we argue, often require a strong reasoning component. Similar to some recent related work, we make visual information accessible to the language model using separate verbalisation models. Specifically, we investigate the performance of open-source, open-access language models against GPT-3 on five vision-language tasks when given textually-encoded visual information. Our results suggest that language models are effective for solving vision-language tasks even with limited samples. This approach also enhances the interpretability of a model's output by providing a means of tracing the output back through the verbalised image content.

Abstract: Training or finetuning large-scale language models (LLMs) such as GPT-3 requires substantial computation resources, motivating recent efforts to explore parameter-efficient adaptation to downstream tasks. One practical area of research is to treat these models as black boxes and interact with them through their inference APIs. In this paper, we investigate how to optimize few-shot text classification without accessing the gradients of the LLMs. To achieve this, we treat the black-box model as a feature extractor and train a classifier with the augmented text data. Data augmentation is performed using prompt-based finetuning on an auxiliary language model with a much smaller parameter size than the black-box model. Through extensive experiments on eight text classification datasets, we show that our approach, dubbed BT-Classifier, significantly outperforms state-of-the-art black-box few-shot learners and performs on par with methods that rely on full-model tuning.

Abstract: Large Language Models (LLMs) have shown stellar achievements in solving a broad range of tasks. When generating text, it is common to sample tokens from these models: whether LLMs closely align with the human disagreement distribution has not been well-studied, especially within the scope of Natural Language Inference (NLI). In this paper, we evaluate the performance and alignment of LLM distribution with humans using two different techniques: Monte Carlo Reconstruction (MCR) and Log Probability Reconstruction (LPR). As a result, we show LLMs exhibit limited ability in solving NLI tasks and simultaneously fail to capture human disagreement distribution, raising concerns about their natural language understanding (NLU) ability and their representativeness of human users.

Abstract: Streaming Automatic Speech Recognition (ASR) in voice assistants can utilize prefetching to partially hide the latency of response generation. Prefetching involves passing a preliminary ASR hypothesis to downstream systems in order to prefetch and cache a response. If the final ASR hypothesis after endpoint detection matches the preliminary one, the cached response can be delivered to the user, thus saving latency. In this paper, we extend this idea by introducing predictive automatic speech recognition, where we predict the full utterance from a partially observed utterance, and prefetch the response based on the predicted utterance. We introduce two personalization approaches and investigate the tradeoff between potential latency gains from successful predictions and the cost increase from failed predictions. We evaluate our methods on an internal voice assistant dataset as well as the public SLURP dataset.

Abstract: The rapid growth of web pages and the increasing complexity of their structure poses a challenge for web mining models. Web mining models are required to understand the semi-structured web pages, particularly when little is known about the subject or template of a new page. Current methods migrate language models to the web mining by embedding the XML source code into the transformer or encoding the rendered layout with graph neural networks. However, these approaches do not take into account the relationships between text nodes within and across pages. In this paper, we propose a new approach, ReXMiner, for zero-shot relation extraction in web mining. ReXMiner encodes the shortest relative paths in the Document Object Model (DOM) tree which is a more accurate and efficient signal for key-value pair extraction within a web page. It also incorporates the popularity of each text node by counting the occurrence of the same text node across different web pages. We use the contrastive learning to address the issue of sparsity in relation extraction. Extensive experiments on public benchmarks show that our method, ReXMiner, outperforms the state-of-the-art baselines in the task of zero-shot relation extraction in web mining.

Abstract: Ambiguous questions persist in open-domain question answering, because formulating a precise question with a unique answer is often challenging. Previously, Min et al. (2020) have tackled this issue by generating disambiguated questions for all possible interpretations of the ambiguous question. This can be effective, but not ideal for providing an answer to the user. Instead, we propose to ask a clarification question, where the user's response will help identify the interpretation that best aligns with the user's intention. We first present CAMBIGNQ, a dataset consisting of 5,654 ambiguous questions, each with relevant passages, possible answers, and a clarification question. The clarification questions were efficiently created by generating them using InstructGPT and manually revising them as necessary. We then define a pipeline of tasks and design appropriate evaluation metrics. Lastly, we achieve 61.3 F1 on ambiguity detection and 40.5 F1 on clarification-based QA, providing strong baselines for future work.

Abstract: Contrastively trained vision-language models have achieved remarkable progress in vision and language representation learning, leading to state-of-the-art models for various downstream multimodal tasks. However, recent research has highlighted severe limitations of these models in their ability to perform compositional reasoning over objects, attributes, and relations. Scene graphs have emerged as an effective way to understand images compositionally. These are graph-structured semantic representations of images that contain objects, their attributes, and relations with other objects in a scene. In this work, we consider the scene graph parsed from text as a proxy for the image scene graph and propose a graph decomposition and augmentation framework along with a coarse-to-fine contrastive learning objective between images and text that aligns sentences of various complexities to the same image. Along with this, we propose novel negative mining techniques in the scene graph space for improving attribute binding and relation understanding. Through extensive experiments, we demonstrate the effectiveness of our approach that significantly improves attribute binding, relation understanding, systematic generalization, and productivity on multiple recently proposed benchmarks (For example, improvements upto $18\%$ for systematic generalization, $16.5\%$ for relation understanding over a strong baseline), while achieving similar or better performance than CLIP on various general multimodal tasks.

Abstract: Objective:Develop and validate an algorithm for analyzing the layout of PDF clinical documents to improve the performance of downstream natural language processing tasks. Materials and Methods: We designed an algorithm to process clinical PDF documents and extract only clinically relevant text. The algorithm consists of several steps: initial text extraction using a PDF parser, followed by classification into categories such as body text, left notes, and footers using a Transformer deep neural network architecture, and finally an aggregation step to compile the lines of a given label in the text. We evaluated the technical performance of the body text extraction algorithm by applying it to a random sample of documents that were annotated. Medical performance was evaluated by examining the extraction of medical concepts of interest from the text in their respective sections. Finally, we tested an end-to-end system on a medical use case of automatic detection of acute infection described in the hospital report. Results:Our algorithm achieved per-line precision, recall, and F1 score of 98.4, 97.0, and 97.7, respectively, for body line extraction. The precision, recall, and F1 score per document for the acute infection detection algorithm were 82.54 (95CI 72.86-91.60), 85.24 (95CI 76.61-93.70), 83.87 (95CI 76, 92-90.08) with exploitation of the results of the advanced body extraction algorithm, respectively. Conclusion:We have developed and validated a system for extracting body text from clinical documents in PDF format by identifying their layout. We were able to demonstrate that this preprocessing allowed us to obtain better performances for a common downstream task, i.e., the extraction of medical concepts in their respective sections, thus proving the interest of this method on a clinical use case.

Abstract: Language identification (LID) is a fundamental step in many natural language processing pipelines. However, current LID systems are far from perfect, particularly on lower-resource languages. We present a LID model which achieves a macro-average F1 score of 0.93 and a false positive rate of 0.033 across 201 languages, outperforming previous work. We achieve this by training on a curated dataset of monolingual data, the reliability of which we ensure by auditing a sample from each source and each language manually. We make both the model and the dataset available to the research community. Finally, we carry out detailed analysis into our model's performance, both in comparison to existing open models and by language class.

Abstract: Conversational implicatures are pragmatic inferences that require listeners to deduce the intended meaning conveyed by a speaker from their explicit utterances. Although such inferential reasoning is fundamental to human communication, recent research indicates that large language models struggle to comprehend these implicatures as effectively as the average human. This paper demonstrates that by incorporating Grice's Four Maxims into the model through chain-of-thought prompting, we can significantly enhance its performance, surpassing even the average human performance on this task.

Abstract: Large language models have demonstrated their ability to self-reflect and refine their generation, which can further improve their performance. However, this feedback mechanism faces challenges such as no guarantee of correctness and the lack of global insight into the model's weaknesses. In this paper, we propose a novel framework, Study Assistant for Large Language Model (SALAM), to aid LLMs in the reflection and refinement process. Motivated by the human study assistant, this framework grades previous responses with the ground truth and collects mistakes in the training phase. During inference, it identifies common misunderstandings based on the mistake collections and provides guidelines for the model to help the model avoid similar mistakes during inference. SALAM is a model-agnostic framework, focusing on providing general feedback and can adapt to any base model. Our evaluation of SALAM on two challenging benchmarks demonstrated a significant improvement over various baselines.

Abstract: Changing speaker names consistently throughout a dialogue should not affect its meaning and corresponding outputs for text generation from dialogues. However, pre-trained language models, serving as the backbone for dialogue-processing tasks, have shown to be sensitive to nuances. This may result in unfairness in real-world applications. No comprehensive analysis of this problem has been done in the past. In this work, we propose to quantitatively measure a model's sensitivity on speaker names, and comprehensively evaluate a number of known methods for reducing speaker name sensitivity, including a novel approach of our own. Extensive experiments on multiple datasets provide a benchmark for this problem and show the favorable performance of our approach in sensitivity reduction and quality of generation.

Abstract: Geoparsing is a fundamental technique for analyzing geo-entity information in text. We focus on document-level geoparsing, which considers geographic relatedness among geo-entity mentions, and presents a Japanese travelogue dataset designed for evaluating document-level geoparsing systems. Our dataset comprises 200 travelogue documents with rich geo-entity information: 12,171 mentions, 6,339 coreference clusters, and 2,551 geo-entities linked to geo-database entries.

Abstract: Visual relation extraction (VRE) aims to extract relations between entities from visuallyrich documents. Existing methods usually predict relations for each entity pair independently based on entity features but ignore the global structure information, i.e., dependencies between entity pairs. The absence of global structure information may make the model struggle to learn long-range relations and easily predict conflicted results. To alleviate such limitations, we propose a GlObal Structure knowledgeguided relation Extraction (GOSE) framework, which captures dependencies between entity pairs in an iterative manner. Given a scanned image of the document, GOSE firstly generates preliminary relation predictions on entity pairs. Secondly, it mines global structure knowledge based on prediction results of the previous iteration and further incorporates global structure knowledge into entity representations. This "generate-capture-incorporate" schema is performed multiple times so that entity representations and global structure knowledge can mutually reinforce each other. Extensive experiments show that GOSE not only outperforms previous methods on the standard fine-tuning setting but also shows promising superiority in cross-lingual learning; even yields stronger data-efficient performance in the low-resource setting.

Abstract: Most task-oriented dialogue (TOD) benchmarks assume users that know exactly how to use the system by constraining the user behaviors within the system's capabilities via strict user goals, namely "user familiarity" bias. This data bias deepens when it combines with data-driven TOD systems, as it is impossible to fathom the effect of it with existing static evaluations. Hence, we conduct an interactive user study to unveil how vulnerable TOD systems are against realistic scenarios. In particular, we compare users with 1) detailed goal instructions that conform to the system boundaries (closed-goal) and 2) vague goal instructions that are often unsupported but realistic (open-goal). Our study reveals that conversations in open-goal settings lead to catastrophic failures of the system, in which 92% of the dialogues had significant issues. Moreover, we conduct a thorough analysis to identify distinctive features between the two settings through error annotation. From this, we discover a novel "pretending" behavior, in which the system pretends to handle the user requests even though they are beyond the system's capabilities. We discuss its characteristics and toxicity while emphasizing transparency and a fallback strategy for robust TOD systems.

Abstract: An outbreak in the popularity of transformer-based Language Models (such as GPT (Brown et al., 2020) and PaLM (Chowdhery et al., 2022)) has opened the doors to new Machine Learning applications. In particular, in Natural Language Processing and how pre-training from large text, corpora is essential in achieving remarkable results in downstream tasks. However, these Language Models seem to have inherent biases toward certain demographics reflected in their training data. While research has attempted to mitigate this problem, existing methods either fail to remove bias altogether, degrade performance, or are expensive. This paper examines the bias produced by promising Language Models when varying parameters and pre-training data. Finally, we propose a de-biasing technique that produces robust de-bias models that maintain performance on downstream tasks.

Abstract: Two fundamental questions in neurolinguistics concerns the brain regions that integrate information beyond the lexical level, and the size of their window of integration. To address these questions we introduce a new approach named masked-attention generation. It uses GPT-2 transformers to generate word embeddings that capture a fixed amount of contextual information. We then tested whether these embeddings could predict fMRI brain activity in humans listening to naturalistic text. The results showed that most of the cortex within the language network is sensitive to contextual information, and that the right hemisphere is more sensitive to longer contexts than the left. Masked-attention generation supports previous analyses of context-sensitivity in the brain, and complements them by quantifying the window size of context integration per voxel.

Abstract: Despite their tremendous successes, most large language models do not have any long-term memory mechanisms, which restricts their applications. Overcoming this limitation would not only require changes to the typical transformer architectures or training procedures, but also a dataset on which these new models could be trained and evaluated. We argue that existing resources lack a few key properties, and that at present, there are no naturalistic datasets of sufficient scale to train (and not only evaluate) long-term memory language models. We then present our solution that capitalizes on the advances in short-term memory language models to create such a dataset. Using GPT 3.5, we summarized each scene in 1500 hand-curated books from Project Gutenberg, which resulted in approximately 150 scene-level summaries per book. We then created a number of reading comprehension questions based on these summaries, including three types of multiple-choice scene recognition questions, as well as free-form narrative reconstruction questions. Each book is thus associated with more than 500 reading comprehension questions. Crucially, most questions have a known ``retention demand'', indicating how long-term of a memory is needed to answer it, which should aid long-term memory performance evaluation. We validate our data in three small-scale experiments: one with human labelers, and two with existing language models. We show that our questions 1) adequately represent the source material 2) can be used to diagnose the model's memory capacity 3) are not trivial for modern language models even when the memory demand does not exceed those models' context lengths. Lastly, we provide our code which can be used to further expand the dataset in an automated manner.

Abstract: While Large Language Models (LLMs) excel in several natural language processing tasks, their size and inaccessibility present challenges for extensive practical application. Previous studies acquire specialized skills through distillation on LLMs, which result in trading generic abilities, called model specialization. As for reasoning ability, chain-of-thought was synthesized to subsequent distillation. However, due to hallucination, synthetic chain-of-thought from LLMs contains faulty reasoning. These incorrect reasoning steps damage the reasoning capability. To tackle above issues, we propose Program-aided Distillation (PaD), which distills LLMs to obtain specialized small models in reasoning tasks. In PaD, we strengthen specialized models with program-aided reasoning, and help them overcome faulty reasoning steps with automated error checking. Experimental results demonstrate that, on the GSM8K benchmark, a 0.06B model using PaD can not only outperform certain LLMs (e.g., LLaMA), but also achieves a 10% improvement over baselines with a significantly smaller scale of parameters and data. Data pruning analysis reveals that PaD possesses higher training efficiency.

Abstract: Despite constituting 65% of all internet traffic in 2023, video content is underrepresented in generative AI research. Meanwhile, recent large language models (LLMs) have become increasingly integrated with capabilities in the visual modality. Integrating video with LLMs is a natural next step, so how can this gap be bridged? To advance video reasoning, we propose a new research direction of VideoCOT on video keyframes, which leverages the multimodal generative abilities of vision-language models to enhance video reasoning while reducing the computational complexity of processing hundreds or thousands of frames. We introduce VIP, an inference-time dataset that can be used to evaluate VideoCOT, containing 1) a variety of real-life videos with keyframes and corresponding unstructured and structured scene descriptions, and 2) two new video reasoning tasks: video infilling and scene prediction. We benchmark various vision-language models on VIP, demonstrating the potential to use vision-language models and LLMs to enhance video chain of thought reasoning.

Abstract: While large language models (LLMs) bring not only performance but also complexity, recent work has started to turn LLMs into data generators rather than task inferencers, where another affordable task model is trained for efficient deployment and inference. However, such an approach has primarily been applied to natural language tasks and has not yet been explored for symbolic language tasks with complex structured outputs (e.g., semantic parsing and code generation). In this paper, we propose SymGen which utilizes LLMs for generating various annotation-expensive symbolic language data. SymGen consists of an informative prompt to steer generation and an agreement-based verifier to improve data correctness. We conduct extensive experiments on six symbolic language tasks across various settings. Compared with the LLMs, we demonstrate the 1\%-sized task model can achieve comparable or better performance, largely cutting inference and deployment costs. We also show that generated data with only a few human demonstrations can be as effective as over 10 times the amount of human-annotated data when training the task model, saving a considerable amount of annotation effort. SymGen sheds new light on data generation for complex tasks, and we release the code at \href{https://github.com/HKUNLP/SymGen}{https://github.com/HKUNLP/SymGen}.

Abstract: The semantic frame induction tasks are defined as a clustering of words into the frames that they evoke, and a clustering of their arguments according to the frame element roles that they should fill. In this paper, we address the latter task of argument clustering, which aims to acquire frame element knowledge, and propose a method that applies deep metric learning. In this method, a pre-trained language model is fine-tuned to be suitable for distinguishing frame element roles through the use of frame-annotated data, and argument clustering is performed with embeddings obtained from the fine-tuned model. Experimental results on FrameNet demonstrate that our method achieves substantially better performance than existing methods.

Abstract: Large Language Models have demonstrated significant ability in accomplishing a wide range of Natural Language Processing (NLP) tasks. However, their performance is highly sensitive to the even minor changes in the phrasing of the task instructions, leading to a line of research in automatic instruction optimization towards better performance for NLP tasks. Unfortunately, existing methods for instruction optimization fail to consider the distribution shift between the seen training data and the unseen test data, where testing on unseen group of data with a different distribution could potentially lead to performance drop. In this paper, we take an initial step of investigating the problem of LLM instruction optimization across data groups with distribution shifts. We find that the optimal instructions do encounter performance drops on LLM under certain distribution shifts. To this end, we propose a framework to derive more robust optimal instructions that improve the performance on the unseen data group without large sacrifice on the seen data group. Experimental results demonstrate the effectiveness of our proposed framework.

Abstract: LLMs have shown impressive few-shot performance across many tasks. However, they still struggle when it comes to generating complex output structures, such as those required for Information Extraction. This limitation stems from the fact that LLMs, without finetuning, tend to generate free text rather than precise structures that follow a specific grammar. In this work, we propose to enrich the decoding step with formal grammar constraints. During beam search, only valid token continuations compliant with the grammar production rules are considered. This enforces the generation of valid sequences exclusively. Our framework is highly general and flexible, allowing any Context-Free Grammar (CFG) to be integrated into our custom constrained beam search implementation. We demonstrate that the outputs of many NLP tasks can be represented as formal languages, making them suitable for direct use in our framework. For task where the output space is dependent on the input, we propose input-dependent grammars to constrain the generation. We conducted experiments with two challenging tasks involving large alphabets in their grammar (Wikidata entities and relations): information extraction and entity disambiguation. Our results with LLaMA models clearly indicate that grammar-constrained decoding outperforms few-shot prompting without constraints, and even competes with task-specific finetuned models. These findings suggest that integrating grammar-based constraints during decoding holds great promise in making LLMs reliably produce structured outputs, especially in setting where training data is scarce and finetuning is expensive.

Abstract: Question answering over knowledge bases (KBQA) aims to answer factoid questions with a given knowledge base (KB). Due to the large scale of KB, annotated data is impossible to cover all fact schemas in KB, which poses a challenge to the generalization ability of methods that require a sufficient amount of annotated data. Recently, LLMs have shown strong few-shot performance in many NLP tasks. We expect LLM can help existing methods improve their generalization ability, especially in low-resource situations. In this paper, we present McL-KBQA, a framework that incorporates the few-shot ability of LLM into the KBQA method via ICL-based multiple choice and then improves the effectiveness of the QA tasks. Experimental results on two KBQA datasets demonstrate the competitive performance of McL-KBQA with strong improvements in generalization. We expect to explore a new way to QA tasks from KBQA in conjunction with LLM, how to generate answers normatively and correctly with strong generalization.

Abstract: Open-domain conversation systems integrate multiple conversation skills into a single system through a modular approach. One of the limitations of the system, however, is the absence of management capability for external memory. In this paper, we propose a simple method to improve BlenderBot3 by integrating memory management ability into it. Since no training data exists for this purpose, we propose an automating dataset creation for memory management. Our method 1) requires little cost for data construction, 2) does not affect performance in other tasks, and 3) reduces external memory. We show that our proposed model BlenderBot3-M^3, which is multi-task trained with memory management, outperforms BlenderBot3 with a relative 4% performance gain in terms of F1 score.

Abstract: The robustness to distribution changes ensures that NLP models can be successfully applied in the realistic world, especially for information extraction tasks. However, most prior evaluation benchmarks have been devoted to validating pairwise matching correctness, ignoring the crucial measurement of robustness. In this paper, we present the first benchmark that simulates the evaluation of open information extraction models in the real world, where the syntactic and expressive distributions under the same knowledge meaning may drift variously. We design and annotate a large-scale testbed in which each example is a knowledge-invariant clique that consists of sentences with structured knowledge of the same meaning but with different syntactic and expressive forms. By further elaborating the robustness metric, a model is judged to be robust if its performance is consistently accurate on the overall cliques. We perform experiments on typical models published in the last decade as well as a popular large language model, the results show that the existing successful models exhibit a frustrating degradation, with a maximum drop of 23.43 F1 score. Our resources and code will be publicly available.

Abstract: In this paper, we present MasakhaPOS, the largest part-of-speech (POS) dataset for 20 typologically diverse African languages. We discuss the challenges in annotating POS for these languages using the UD (universal dependencies) guidelines. We conducted extensive POS baseline experiments using conditional random field and several multilingual pre-trained language models. We applied various cross-lingual transfer models trained with data available in UD. Evaluating on the MasakhaPOS dataset, we show that choosing the best transfer language(s) in both single-source and multi-source setups greatly improves the POS tagging performance of the target languages, in particular when combined with cross-lingual parameter-efficient fine-tuning methods. Crucially, transferring knowledge from a language that matches the language family and morphosyntactic properties seems more effective for POS tagging in unseen languages.

Abstract: Incorporating language-specific (LS) modules is a proven method to boost performance in multilingual machine translation. This approach bears similarity to Mixture-of-Experts (MoE) because it does not inflate FLOPs. However, the scalability of this approach to hundreds of languages (experts) tends to be unmanageable due to the prohibitive number of parameters introduced by full-rank matrices in fully-connected layers. In this work, we introduce the Language-Specific Matrix Synthesis (LMS) method. This approach constructs LS modules by generating low-rank matrices from two significantly smaller matrices to approximate the full-rank matrix. Furthermore, we condense multilingual knowledge from multiple LS modules into a single shared module with the Fuse Distillation (FD) technique to improve the efficiency of inference and model serialization. We show that our LMS method significantly outperforms previous LS methods and MoE methods with the same amount of extra parameters, e.g., 1.73 BLEU points over the Switch Transformer on many-to-many multilingual machine translation. Importantly, LMS is able to have comparable translation performance with much fewer parameters.

Abstract: Large and sparse feed-forward networks (S-FFN) such as Mixture-of-Experts (MoE) have demonstrated to be an efficient approach for scaling up Transformers model size for pretraining large language models. By only activating part of the FFN parameters conditioning on input, S-FFN improves generalization performance while keeping training and inference costs (in FLOPs) fixed. In this work, we analyzed the two major design choices of S-FFN: the memory block (or expert) size and the memory block selection method under a general conceptual framework of sparse neural memory. Using this unified framework, we compare several S-FFN architectures for language modeling and provide insights into their relative efficacy and efficiency. From our analysis results, we found a simpler selection method -- Avg-K that selects blocks through their mean aggregated hidden states, achieves lower perplexity in language modeling pretraining compared to existing MoE architectures.

Abstract: Large language models (LLMs) exhibit remarkable performance across various NLP tasks. However, they often generate incorrect or hallucinated information, which hinders their practical applicability in real-world scenarios. Human feedback has been shown to effectively enhance the factuality and quality of generated content, addressing some of these limitations. However, this approach is resource-intensive, involving manual input and supervision, which can be time-consuming and expensive. Moreover, it cannot be provided during inference, further limiting its practical utility in dynamic and interactive applications. In this paper, we introduce ReFeed, a novel pipeline designed to enhance LLMs by providing automatic retrieval feedback in a plug-and-play framework without the need for expensive fine-tuning. ReFeed first generates initial outputs, then utilizes a retrieval model to acquire relevant information from large document collections, and finally incorporates the retrieved information into the in-context demonstration for output refinement, thereby addressing the limitations of LLMs in a more efficient and cost-effective manner. Experiments on four knowledge-intensive benchmark datasets demonstrate our proposed ReFeed could improve over +6.0% under zero-shot setting and +2.5% under few-shot setting, compared to baselines without using retrieval feedback.

Abstract: Sanskrit is a low-resource language with a rich heritage. Digitized Sanskrit corpora reflective of the contemporary usage of Sanskrit, specifically that too in prose, is heavily under-represented at present. Presently, no such English-Sanskrit parallel dataset is publicly available. We release a dataset, S\={a}mayik, of more than 42,000 parallel English-Sanskrit sentences, from four different corpora that aim to bridge this gap. Moreover, we also release benchmarks adapted from existing multilingual pretrained models for Sanskrit-English translation. We include training splits from our contemporary dataset and the Sanskrit-English parallel sentences from the training split of Itih\={a}sa, a previously released classical era machine translation dataset containing Sanskrit.

Abstract: The goal of dialogue topic shift detection is to identify whether the current topic in a conversation has changed or needs to change. Previous work focused on detecting topic shifts using pre-trained models to encode the utterance, failing to delve into the various levels of topic granularity in the dialogue and understand dialogue contents. To address the above issues, we take a prompt-based approach to fully extract topic information from dialogues at multiple-granularity, i.e., label, turn, and topic. Experimental results on our annotated Chinese Natural Topic Dialogue dataset CNTD and the publicly available English TIAGE dataset show that the proposed model outperforms the baselines. Further experiments show that the information extracted at different levels of granularity effectively helps the model comprehend the conversation topics.

Abstract: Multi-task learning (MTL) aims at achieving a better model by leveraging data and knowledge from multiple tasks. However, MTL does not always work -- sometimes negative transfer occurs between tasks, especially when aggregating loosely related skills, leaving it an open question when MTL works. Previous studies show that MTL performance can be improved by algorithmic tricks. However, what tasks and skills should be included is less well explored. In this work, we conduct a case study in Financial NLP where multiple datasets exist for skills relevant to the domain, such as numeric reasoning and sentiment analysis. Due to the task difficulty and data scarcity in the Financial NLP domain, we explore when aggregating such diverse skills from multiple datasets with MTL can work. Our findings suggest that the key to MTL success lies in skill diversity, relatedness between tasks, and choice of aggregation size and shared capacity. Specifically, MTL works well when tasks are diverse but related, and when the size of the task aggregation and the shared capacity of the model are balanced to avoid overwhelming certain tasks.

Abstract: Although counterfactual reasoning is a fundamental aspect of intelligence, the lack of large-scale counterfactual open-domain question-answering (QA) benchmarks makes it difficult to evaluate and improve models on this ability. To address this void, we introduce the first such dataset, named IfQA, where each question is based on a counterfactual presupposition via an "if" clause. For example, if Los Angeles was on the east coast of the U.S., what would be the time difference between Los Angeles and Paris? Such questions require models to go beyond retrieving direct factual knowledge from the Web: they must identify the right information to retrieve and reason about an imagined situation that may even go against the facts built into their parameters. The IfQA dataset contains over 3,800 questions that were annotated annotated by crowdworkers on relevant Wikipedia passages. Empirical analysis reveals that the IfQA dataset is highly challenging for existing open-domain QA methods, including supervised retrieve-then-read pipeline methods (EM score 36.2), as well as recent few-shot approaches such as chain-of-thought prompting with GPT-3 (EM score 27.4). The unique challenges posed by the IfQA benchmark will push open-domain QA research on both retrieval and counterfactual reasoning fronts.

Abstract: Existing approaches for unsupervised bilingual lexicon induction (BLI) often depend on good quality static or contextual embeddings trained on large monolingual corpora for both languages. In reality, however, unsupervised BLI is most likely to be useful for dialects and languages that do not have abundant amounts of monolingual data. We introduce a simple and fast method for unsupervised BLI for low-resource languages with a related mid-to-high resource language, only requiring inference on the higher-resource language monolingual BERT. We work with two low-resource languages ($<5M$ monolingual tokens), Bhojpuri and Magahi, of the severely under-researched Indic dialect continuum, showing that state-of-the-art methods in the literature show near-zero performance in these settings, and that our simpler method gives much better results. We repeat our experiments on Marathi and Nepali, two higher-resource Indic languages, to compare approach performances by resource range. We release automatically created bilingual lexicons for the first time for five languages of the Indic dialect continuum.

Abstract: Gender bias is a significant issue in machine translation, leading to ongoing research efforts in developing bias mitigation techniques. However, most works focus on debiasing of bilingual models without consideration for multilingual systems. In this paper, we specifically target the unambiguous gender bias issue of multilingual machine translation models and propose a new mitigation method based on a novel perspective on the problem. We hypothesize that the gender bias in unambiguous settings is due to the lack of gender information encoded into the non-explicit gender words and devise a scheme to encode correct gender information into their latent embeddings. Specifically, we employ Gender-Aware Contrastive Learning, GACL, based on gender pseudo-labels to encode gender information on the encoder embeddings. Our method is target-language-agnostic and applicable to already trained multilingual machine translation models through post-fine-tuning. Through multilingual evaluation, we show that our approach improves gender accuracy by a wide margin without hampering translation performance. We also observe that incorporated gender information transfers and benefits other target languages regarding gender accuracy. Finally, we demonstrate that our method is applicable and beneficial to models of various sizes.

Abstract: ``Theory of Mind" (ToM) is the ability to understand human thinking and decision-making, an ability that plays a crucial role in many types of social interaction between people, including linguistic communication. This paper investigates to what extent recent Large Language Models in the ChatGPT tradition possess ToM. Focussing on six well-known ToM problems, we posed each problem to two versions of ChatGPT and compared the results under a range of prompting strategies. While the results concerning ChatGPT-3 were somewhat inconclusive, ChatGPT-4 was shown to arrive at the correct answers more often than would be expected based on chance, although correct answers were often arrived at on the basis of false assumptions or invalid reasoning.

Abstract: In speech translation, leveraging multimodal data to improve model performance and address limitations of individual modalities has shown significant effectiveness. In this paper, we harness the complementary strengths of speech and text, which are disparate modalities. We observe three levels of modality gap between them, denoted by Modal input representation, Modal semantic, and Modal hidden states. To tackle these gaps, we propose \textbf{F}use-\textbf{S}peech-\textbf{T}ext (\textbf{FST}), a cross-modal model which supports three distinct input modalities for translation: speech, text, and fused speech-text. We leverage multiple techniques for cross-modal alignment and conduct a comprehensive analysis to assess its impact on speech translation, machine translation, and fused speech-text translation. We evaluate FST on MuST-C, GigaST, and newstest benchmark. Experiments show that the proposed FST achieves an average 34.0 BLEU on MuST-C En$\rightarrow$De/Es/Fr (vs SOTA +1.1 BLEU). Further experiments demonstrate that FST does not degrade on MT task, as observed in prior works. Instead, it yields an average improvement of 3.2 BLEU over the pre-trained MT model.

Abstract: In this paper we present the Process-To-Text (P2T) framework for the automatic generation of textual descriptive explanations of processes. P2T integrates three AI paradigms: process mining for extracting temporal and structural information from a process, fuzzy linguistic protoforms for modelling uncertain terms, and natural language generation for building the explanations. A real use-case in the cardiology domain is presented, showing the potential of P2T for providing natural language explanations addressed to specialists.

Abstract: Large Language Models (LLMs) have shown enhanced capabilities of solving novel tasks by reasoning step-by-step known as Chain-of-Thought (CoT) reasoning; how can we instill the same capability of reasoning step-by-step on unseen tasks into LMs that possess less than <100B parameters? To address this question, we first introduce the CoT Collection, a new instruction-tuning dataset that augments 1.88 million CoT rationales across 1,060 tasks. We show that continually fine-tuning Flan-T5 (3B & 11B) with the CoT Collection enables the 3B & 11B LMs to perform CoT better on unseen tasks, leading to an improvement in the average zero-shot accuracy on 27 datasets of the BIG-Bench-Hard benchmark by +4.34% and +2.44%, respectively. Furthermore, we show that instruction tuning with CoT allows LMs to possess stronger few-shot learning capabilities, resulting in an improvement of +2.97% and +2.37% on 4 domain-specific tasks over Flan-T5 (3B & 11B), respectively. We make our CoT Collection data and our trained models publicly available at https://github.com/kaist-lklab/CoT-Collection.

Abstract: Attention-based encoder-decoder (AED) models have shown impressive performance in ASR. However, most existing AED methods neglect to simultaneously leverage both acoustic and semantic features in decoder, which is crucial for generating more accurate and informative semantic states. In this paper, we propose an Acoustic and Semantic Cooperative Decoder (ASCD) for ASR. In particular, unlike vanilla decoders that process acoustic and semantic features in two separate stages, ASCD integrates them cooperatively. To prevent information leakage during training, we design a Causal Multimodal Mask. Moreover, a variant Semi-ASCD is proposed to balance accuracy and computational cost. Our proposal is evaluated on the publicly available AISHELL-1 and aidatatang_200zh datasets using Transformer, Conformer, and Branchformer as encoders, respectively. The experimental results show that ASCD significantly improves the performance by leveraging both the acoustic and semantic information cooperatively.

Abstract: Language models~(LMs) gradually become general-purpose interfaces in the interactive and embodied world, where the understanding of physical concepts is an essential prerequisite. However, it is not yet clear whether LMs can understand physical concepts in the human world. To investigate this, we design a benchmark VEC that covers the tasks of (i) Visual concepts, such as the shape and material of objects, and (ii) Embodied Concepts, learned from the interaction with the world such as the temperature of objects. Our zero (few)-shot prompting results show that the understanding of certain visual concepts emerges as scaling up LMs, but there are still basic concepts to which the scaling law does not apply. For example, OPT-175B performs close to humans with a zero-shot accuracy of 85\% on the material concept, yet behaves like random guessing on the mass concept. Instead, vision-augmented LMs such as CLIP and BLIP achieve a human-level understanding of embodied concepts. Analysis indicates that the rich semantics in visual representation can serve as a valuable source of embodied knowledge. Inspired by this, we propose a distillation method to transfer embodied knowledge from VLMs to LMs, achieving performance gain comparable with that by scaling up the parameters of LMs 134x. Our dataset is available at \url{https://github.com/TobiasLee/VEC}

Abstract: Training models with varying capacities can be advantageous for deploying them in different scenarios. While high-capacity models offer better performance, low-capacity models require fewer computing resources for training and inference. In this work, we propose a novel one-stop training framework consisting of two composite model architectures and a joint training algorithm called Two-Stage Joint-Training (TSJT). Unlike knowledge distillation, where multiple capacity models are trained from scratch separately, our approach integrates supervisions from different flexible-capacity models simultaneously, leading to faster and more efficient convergence. Extensive experiments on the WMT10 benchmark show that our method outperforms low-capacity baseline models and achieves comparable or better performance on high-capacity models. Notably, the analysis demonstrates that our method significantly influences the initial training process, leading to more efficient convergence and superior solutions.

Abstract: Detecting factual errors in summaries has been an important and challenging subject in summarization research. Inspired by the emergent ability of large language models (LLMs), we explore evaluating factual consistency of summaries by directly prompting LLMs. We present a comprehensive empirical study to assess the ability of LLMs as factual consistency evaluators, which consists of (1) analyzing different LLMs such as the GPT model series and Flan-T5; (2) investigating a variety of prompting methods including vanilla prompting, chain-of-thought prompting, and a sentence-by-sentence prompting method to tackle long summaries; and (3) evaluating on diverse summaries generated by multiple summarization systems, ranging from pre-transformer methods to SOTA pretrained models. Our experiments demonstrate that prompting LLMs is able to outperform the previous best factuality systems in all settings, by up to 12.2 absolute points in terms of the binary classification accuracy on inconsistency detection.

Abstract: Much recent effort has been devoted to creating large-scale language models. Nowadays, the most prominent approaches are based on deep neural networks, such as BERT. However, they lack transparency and interpretability, and are often seen as black boxes. This affects not only their applicability in downstream tasks but also the comparability of different architectures or even of the same model trained using different corpora or hyperparameters. In this paper, we propose a set of intrinsic evaluation tasks that inspect the linguistic information encoded in models developed for Brazilian Portuguese. These tasks are designed to evaluate how different language models generalise information related to grammatical structures and multiword expressions (MWEs), thus allowing for an assessment of whether the model has learned different linguistic phenomena. The dataset that was developed for these tasks is composed of a series of sentences with a single masked word and a cue phrase that helps in narrowing down the context. This dataset is divided into MWEs and grammatical structures, and the latter is subdivided into 6 tasks: impersonal verbs, subject agreement, verb agreement, nominal agreement, passive and connectors. The subset for MWEs was used to test BERTimbau Large, BERTimbau Base and mBERT. For the grammatical structures, we used only BERTimbau Large, because it yielded the best results in the MWE task.

Abstract: In Emotion Recognition in Conversations (ERC), the emotions of target utterances are closely dependent on their context. Therefore, existing works train the model to generate the response of the target utterance, which aims to recognise emotions leveraging contextual information. However, adjacent response generation ignores long-range dependencies and provides limited affective information in many cases. In addition, most ERC models learn a unified distributed representation for each utterance, which lacks interpretability and robustness. To address these issues, we propose a VAD-disentangled Variational AutoEncoder (VAD-VAE), which first introduces a target utterance reconstruction task based on Variational Autoencoder, then disentangles three affect representations Valence-Arousal-Dominance (VAD) from the latent space. We also enhance the disentangled representations by introducing VAD supervision signals from a sentiment lexicon and minimising the mutual information between VAD distributions. Experiments show that VAD-VAE outperforms the state-of-the-art model on two datasets. Further analysis proves the effectiveness of each proposed module and the quality of disentangled VAD representations. The code is available at https://github.com/SteveKGYang/VAD-VAE.

Abstract: Due to the broad range of social media platforms and their user groups, the requirements of abusive language detection systems are varied and ever-changing. Already a large set of annotated corpora with different properties and label sets were created, such as hate or misogyny detection, but the form and targets of abusive speech are constantly changing. Since, the annotation of new corpora is expensive, in this work we leverage datasets we already have, covering a wide range of tasks related to abusive language detection, in order to build models cheaply for a new target label set and/or language, using only a few training examples of the target domain. We propose a two-step approach: first we train our model in a multitask fashion. We then carry out few-shot adaptation to the target requirements. Our experiments show that by leveraging already existing datasets and only a few-shots of the target task the performance of models can be improved not only monolingually but across languages as well. Our analysis also shows that our models acquire a general understanding of abusive language, since they improve the prediction of labels which are present only in the target dataset. We also analyze the trade-off between specializing the already existing datasets to a given target setup for best performance and its negative effects on model adaptability.

Abstract: Given BM25's enduring competitiveness as an information retrieval baseline, we investigate to what extent it can be even further improved by augmenting and re-weighting its sparse query-vector representation. We propose an approach to learning an augmentation and a re-weighting end-to-end, and we find that our approach improves performance over BM25 while retaining its speed. We furthermore find that the learned augmentations and re-weightings transfer well to unseen datasets.

Abstract: Years have passed since the NLP community has last focused on linguistic acceptability. In this work, we revisit this topic in the context of large language models. We introduce CoLAC - Corpus of Linguistic Acceptability in Chinese, the first large-scale non-English acceptability dataset that is verified by native speakers and comes with two sets of labels. Our experiments show that even the largest InstructGPT model performs only at chance level on CoLAC, while ChatGPT's performance (48.30 MCC) is also way below supervised models (59.03 MCC) and human (65.11 MCC). Through cross-lingual transfer experiments and fine-grained linguistic analysis, we demonstrate for the first time that knowledge of linguistic acceptability can be transferred across typologically distinct languages, as well as be traced back to pre-training.

Abstract: Machine learning (ML) systems in natural language processing (NLP) face significant challenges in generalizing to out-of-distribution (OOD) data, where the test distribution differs from the training data distribution. This poses important questions about the robustness of NLP models and their high accuracy, which may be artificially inflated due to their underlying sensitivity to systematic biases. Despite these challenges, there is a lack of comprehensive surveys on the generalization challenge from an OOD perspective in text classification. Therefore, this paper aims to fill this gap by presenting the first comprehensive review of recent progress, methods, and evaluations on this topic. We furth discuss the challenges involved and potential future research directions. By providing quick access to existing work, we hope this survey will encourage future research in this area.

Abstract: Large language models have demonstrated the capability to perform well on many NLP tasks when the input is prompted with a few examples (in-context learning) including machine translation, which is the focus of this work. The quality of translation depends on various features of the selected examples, such as their quality and relevance. However, previous work has predominantly focused on individual features for example selection. We propose a general framework for combining different features influencing example selection. We learn a regression function that selects examples based on multiple features in order to maximize the translation quality. On multiple language pairs and language models, we show that our example selection method significantly outperforms random selection as well as strong single-factor baselines reported in the literature. Using our example selection method, we see an improvement of over 2.5 COMET points on average with respect to a strong BM25 retrieval-based baseline.

Abstract: Modern large language models (LLMs) have demonstrated impressive capabilities at sophisticated tasks, often through step-by-step reasoning similar to humans. This is made possible by their strong few and zero-shot abilities -- they can effectively learn from a handful of handcrafted, completed responses ("in-context examples"), or are prompted to reason spontaneously through specially designed triggers. Nonetheless, some limitations have been observed. First, performance in the few-shot setting is sensitive to the choice of examples, whose design requires significant human effort. Moreover, given the diverse downstream tasks of LLMs, it may be difficult or laborious to handcraft per-task labels. Second, while the zero-shot setting does not require handcrafting, its performance is limited due to the lack of guidance to the LLMs. To address these limitations, we propose Consistency-based Self-adaptive Prompting (COSP), a novel prompt design method for LLMs. Requiring neither handcrafted responses nor ground-truth labels, COSP selects and builds the set of examples from the LLM zero-shot outputs via carefully designed criteria that combine consistency, diversity and repetition. In the zero-shot setting for three different LLMs, we show that using only LLM predictions, COSP improves performance up to 15% compared to zero-shot baselines and matches or exceeds few-shot baselines for a range of reasoning tasks.

Abstract: Multilingual machine translation (MMT), trained on a mixture of parallel and monolingual data, is key for improving translation in low-resource language pairs. However, the literature offers conflicting results on the performance of different methods. To resolve this, we examine how denoising autoencoding (DAE) and backtranslation (BT) impact MMT under different data conditions and model scales. Unlike prior studies, we use a realistic dataset of 100 directions and consider many domain combinations of monolingual and test data. We find that monolingual data generally helps MMT, but models are surprisingly brittle to domain mismatches, especially at smaller model scales. BT is beneficial when the parallel, monolingual, and test data sources are similar but can be detrimental otherwise, while DAE is less effective than previously reported. Next, we analyze the impact of scale (from 90M to 1.6B parameters) and find it is important for both methods, particularly DAE. As scale increases, DAE transitions from underperforming the parallel-only baseline at 90M to converging with BT performance at 1.6B, and even surpassing it in low-resource. These results offer new insights into how to best use monolingual data in MMT.

Abstract: Embedding models have shown great power in knowledge graph completion (KGC) task. By learning structural constraints for each training triple, these methods implicitly memorize intrinsic relation rules to infer missing links. However, this paper points out that the multi-hop relation rules are hard to be reliably memorized due to the inherent deficiencies of such implicit memorization strategy, making embedding models underperform in predicting links between distant entity pairs. To alleviate this problem, we present Vertical Learning Paradigm (VLP), which extends embedding models by allowing to explicitly copy target information from related factual triples for more accurate prediction. Rather than solely relying on the implicit memory, VLP directly provides additional cues to improve the generalization ability of embedding models, especially making the distant link prediction significantly easier. Moreover, we also propose a novel relative distance based negative sampling technique (ReD) for more effective optimization. Experiments demonstrate the validity and generality of our proposals on two standard benchmarks. Our code is available at https://github.com/rui9812/VLP.

Abstract: In-context learning (ICL), teaching a large language model (LLM) to perform a task with few-shot demonstrations rather than adjusting the model parameters, has emerged as a strong paradigm for using LLMs. While early studies primarily used a fixed or random set of demonstrations for all test queries, recent research suggests that retrieving semantically similar demonstrations to the input from a pool of available demonstrations results in better performance. This work expands the applicability of retrieval-based ICL approaches by demonstrating that even simple word-overlap similarity measures such as BM25 outperform randomly selected demonstrations. Furthermore, we extend the success of retrieval-based ICL to instruction-finetuned LLMs as well as Chain-of-Thought (CoT) prompting. For instruction-finetuned LLMs, we find that although a model has already seen the training data at training time, retrieving demonstrations from the training data at test time yields better results compared to using no demonstrations or random demonstrations. Last but not least, we train a task-specific demonstration retriever that outperforms off-the-shelf retrievers.

Abstract: To fully evaluate the overall performance of different NLP models in a given domain, many evaluation benchmarks are proposed, such as GLUE, SuperGLUE and CLUE. The fi eld of natural language understanding has traditionally focused on benchmarks for various tasks in languages such as Chinese, English, and multilingua, however, there has been a lack of attention given to the area of classical Chinese, also known as "wen yan wen", which has a rich history spanning thousands of years and holds signifi cant cultural and academic value. For the prosperity of the NLP community, in this paper, we introduce the WYWEB evaluation benchmark, which consists of nine NLP tasks in classical Chinese, implementing sentence classifi cation, sequence labeling, reading comprehension, and machine translation. We evaluate the existing pre-trained language models, which are all struggling with this benchmark. We also introduce a number of supplementary datasets and additional tools to help facilitate further progress on classical Chinese NLU. The github repository is https://github.com/baudzhou/WYWEB.

Abstract: In-context learning (ICL) emerges as a promising capability of large language models (LLMs) by providing them with demonstration examples to perform diverse tasks. However, the underlying mechanism of how LLMs learn from the provided context remains under-explored. In this paper, we investigate the working mechanism of ICL through an information flow lens. Our findings reveal that label words in the demonstration examples function as anchors: (1) semantic information aggregates into label word representations during the shallow computation layers' processing; (2) the consolidated information in label words serves as a reference for LLMs' final predictions. Based on these insights, we introduce an anchor re-weighting method to improve ICL performance, a demonstration compression technique to expedite inference, and an analysis framework for diagnosing ICL errors in GPT2-XL. The promising applications of our findings again validate the uncovered ICL working mechanism and pave the way for future studies.

Abstract: Event detection is a crucial information extraction task in many domains, such as Wikipedia or news. The task typically relies on trigger detection (TD) -- identifying token spans in the text that evoke specific events. While the notion of triggers should ideally be universal across domains, domain transfer for TD from high- to low-resource domains results in significant performance drops. We address the problem of negative transfer for TD by coupling triggers between domains using subject-object relations obtained from a rule-based open information extraction (OIE) system. We demonstrate that relations injected through multi-task training can act as mediators between triggers in different domains, enhancing zero- and few-shot TD domain transfer and reducing negative transfer, in particular when transferring from a high-resource source Wikipedia domain to a low-resource target news domain. Additionally, we combine the extracted relations with masked language modeling on the target domain and obtain further TD performance gains. Finally, we demonstrate that the results are robust to the choice of the OIE system.

Abstract: Large language models are able to learn new tasks in context, where they are provided with instructions and a few annotated examples. However, the effectiveness of in-context learning is dependent to the provided context, and the performance on a downstream task can vary a lot depending on the instruction. Importantly, such dependency on the context can happen in unpredictable ways, e.g., a seemingly more informative instruction might lead to a worse performance. In this paper, we propose an alternative approach, which we term in-context probing. Similar to in-context learning, we contextualize the representation of the input with an instruction, but instead of decoding the output prediction, we probe the contextualized representation to predict the label. Through a series of experiments on a diverse set of classification tasks, we show that in-context probing is significantly more robust to changes in instructions. We further show that probing can be particularly helpful to build classifiers on top of smaller models, and with only a hundred training examples.

Abstract: Using a shared vocabulary is common practice in Multilingual Neural Machine Translation (MNMT). In addition to its simple design, shared tokens play an important role in positive knowledge transfer, which manifests naturally when the shared tokens refer to similar meanings across languages. However, natural flaws exist in such a design as well: 1) when languages use different writing systems, transfer is inhibited, and 2) even if languages use similar writing systems, shared tokens may have completely different meanings in different languages, increasing ambiguity. In this paper, we propose a re-parameterized method for building embeddings to alleviate the first problem. More specifically, we define word-level information transfer pathways via word equivalence classes and rely on graph networks to fuse word embeddings across languages. Our experiments demonstrate the advantages of our approach: 1) the semantics of embeddings are better aligned across languages, 2) our method achieves significant BLEU improvements on high- and low-resource MNMT, and 3) only less than 1.0\% additional trainable parameters are required with a limited increase in computational costs.

Abstract: While large pre-trained language models (LMs) find greater use across NLP, existing evaluation protocols do not consider how LM language use aligns with particular human demographic groups, which can be an important consideration in conversational AI applications. To remedy this gap, we consider how LM language skills can be measured and compared to human sub-populations. We suggest clinical techniques from Speech Language Pathology, which has well-established norms for acquisition of language skills, organized by (human) age. We conduct evaluation with a domain expert (i.e., a clinically licensed speech language pathologist), and also propose automated techniques to substitute clinical evaluation at scale. We find LM capability varies widely depending on task with GPT-3.5 mimicking the ability of a typical 6-9 year old at tasks requiring inference about word meanings and simultaneously outperforming a typical 21 year old at memorization. GPT-3.5 (InstructGPT) also has trouble with social language use, exhibiting less than 50\% of the tested pragmatic skills. It shows errors in understanding particular word parts-of-speech and associative word relations, among other lexical features. Ultimately, findings reiterate the importance of considering demographic alignment and conversational goals when using these models as public-facing tools. Our framework will be publicly available via code, data, and a python package.

Abstract: We introduce ZeroSCROLLS, a zero-shot benchmark for natural language understanding over long texts, which contains only test sets, without training or development data. We adapt six tasks from the SCROLLS benchmark, and add four new datasets, including two novel information fusing tasks, such as aggregating the percentage of positive reviews. Using ZeroSCROLLS, we conduct a comprehensive evaluation of both open-source and closed large language models, finding that Claude outperforms ChatGPT, and that GPT-4 achieves the highest average score. However, there is still room for improvement on multiple open challenges in ZeroSCROLLS, such as aggregation tasks, where models struggle to pass the naive baseline. As the state of the art is a moving target, we invite researchers to evaluate their ideas on the live ZeroSCROLLS leaderboard

Abstract: The striking ability of unsupervised word translation has been demonstrated with the help of word vectors / pretraining; however, they require large amounts of data and usually fails if the data come from different domains. We propose coocmap, a method that can use either high-dimensional co-occurrence counts or their lower-dimensional approximations. Freed from the limits of low dimensions, we show that relying on low-dimensional vectors and their incidental properties miss out on better denoising methods and useful world knowledge in high dimensions, thus stunting the potential of the data. Our results show that unsupervised translation can be achieved more easily and robustly than previously thought -- less than 80MB and minutes of CPU time is required to achieve over 50\% accuracy for English to Finnish, Hungarian, and Chinese translations when trained on similar data; even under domain mismatch, we show coocmap still works fully unsupervised on English NewsCrawl to Chinese Wikipedia and English Europarl to Spanish Wikipedia, among others. These results challenge prevailing assumptions on the necessity and superiority of low-dimensional vectors, and suggest that similarly processed co-occurrences can outperform dense vectors on other tasks too.

Abstract: As the largest knowledge base, Wikidata is a massive source of knowledge, complementing large language models with well-structured data. In this paper, we present WikiWebQuestions, a high-quality knowledge base question answering benchmark for Wikidata. This new benchmark uses real-world human data with SPARQL annotation to facilitate a more accurate comparison with large language models utilizing the up-to-date answers from Wikidata. Additionally, a baseline for this benchmark is established with an effective training data synthesis methodology and WikiSP, a Seq2Seq semantic parser, that handles large noisy knowledge graphs. Experimental results illustrate the effectiveness of this methodology, achieving 69% and 59% answer accuracy in the dev set and test set, respectively. We showed that we can pair semantic parsers with GPT-3 to provide a combination of verifiable results and qualified guesses that can provide useful answers to 97% of the questions in the dev set of our benchmark.

Abstract: Cross-lingual summarization consists of generating a summary in one language given an input document in a different language, allowing for the dissemination of relevant content across speakers of other languages. However, this task remains challenging, mainly because of the need for cross-lingual datasets and the compounded difficulty of summarizing and translating. This work presents $\mu$PLAN, an approach to cross-lingual summarization that uses an intermediate planning step as a cross-lingual bridge. We formulate the plan as a sequence of entities that captures the conceptualization of the summary, i.e. identifying the salient content and expressing in which order to present the information, separate from the surface form. Using a multilingual knowledge base, we align the entities to their canonical designation across languages. $\mu$PLAN models first learn to generate the plan and then continue generating the summary conditioned on the plan and the input. We evaluate our methodology on the XWikis dataset on cross-lingual pairs across four languages and demonstrate that this planning objective achieves state-of-the-art performance in terms of ROUGE and faithfulness scores. Moreover, this planning approach improves the zero-shot transfer to new cross-lingual language pairs compared to non-planning baselines.

Abstract: Large, general purpose language models have demonstrated impressive performance across many different conversational domains. While multi-domain language models achieve low overall perplexity, their outputs are not guaranteed to stay within the domain of a given input prompt. This paper proposes domain privacy as a novel way to quantify how likely a conditional language model will leak across domains. We also develop policy functions based on token-level domain classification, and propose an efficient fine-tuning method to improve the trained model's domain privacy. Experiments on membership inference attacks show that our proposed method has comparable resiliency to methods adapted from recent literature on differentially private language models.

Abstract: In-Context learning is the paradigm that adapts large language models to downstream tasks by providing a few examples. Few-shot selection -- selecting appropriate examples for each test instance separately -- is important for in-context learning. In this paper, we propose Skill-KNN, a skill-based few-shot selection method for in-context learning. The key advantages of Skill-KNN include: (1) it addresses the problem that existing methods based on pre-trained embeddings can be easily biased by surface natural language features that are not important for the target task; (2) it does not require training or fine-tuning of any models, making it suitable for frequently expanding or changing example banks. The key insight is to optimize the inputs fed into the embedding model, rather than tuning the model itself. Technically, Skill-KNN generates the skill-based representations for each test case and candidate example by utilizing a pre-processing few-shot prompting, thus eliminating unimportant surface features. Experimental results across four cross-domain semantic parsing tasks and four backbone models show that Skill-KNN significantly outperforms existing methods.

Abstract: In Textual question answering (TQA) systems, complex questions often require retrieving multiple textual fact chains with multiple reasoning steps. While existing benchmarks are limited to single-chain or single-hop retrieval scenarios. In this paper, we propose to conduct Graph-Hop -- a novel multi-chains and multi-hops retrieval and reasoning paradigm in complex question answering. We construct a new benchmark called ReasonGraphQA, which provides explicit and fine-grained evidence graphs for complex questions to support interpretable reasoning, comprehensive and detailed reasoning. And ReasonGraphQA also shows an advantage in reasoning diversity and scale. Moreover, We propose a strong graph-hop baseline called Bidirectional Graph Retrieval (BGR) method for generating an explanation graph of textual evidence in knowledge reasoning and question answering. We have thoroughly evaluated existing evidence retrieval and reasoning models on the ReasonGraphQA. Experiments highlight Graph-Hop is a promising direction for answering complex questions, but it still has certain limitations. We have further studied mitigation strategies to meet these challenges and discuss future directions.

Abstract: While many languages possess processes of joining two or more words to create compound words, previous studies have been typically limited only to languages with excessively productive compound formation (e.g., German, Dutch) and there is no public dataset containing compound and non-compound words across a large number of languages. In this work, we systematically study decompounding, the task of splitting compound words into their constituents, at a wide scale. We first address the data gap by introducing a dataset of 255k compound and non-compound words across 56 diverse languages obtained from Wiktionary. We then use this dataset to evaluate an array of Large Language Models (LLMs) on the decompounding task. We find that LLMs perform poorly, especially on words which are tokenized unfavorably by subword tokenization. We thus introduce a novel methodology to train dedicated models for decompounding. The proposed two-stage procedure relies on a fully self-supervised objective in the first stage, while the second, supervised learning stage optionally fine-tunes the model on the annotated Wiktionary data. Our self-supervised models outperform the prior best unsupervised decompounding models by 13.9% accuracy on average. Our fine-tuned models outperform all prior (language-specific) decompounding tools. Furthermore, we use our models to leverage decompounding during the creation of a subword tokenizer, which we refer to as CompoundPiece. CompoundPiece tokenizes compound words more favorably on average, leading to improved performance on decompounding over an otherwise equivalent model using SentencePiece tokenization.

Abstract: Conventional supervised approaches for text-to-SQL parsing often require large amounts of annotated data, which is costly to obtain in practice. Recently, in-context learning with large language models (LLMs) has caught increasing attention due to its superior few-shot performance in a wide range of tasks. However, most attempts to use in-context learning for text-to-SQL parsing still lag behind supervised methods. We hypothesize that the under-performance is because text-to-SQL parsing requires complex, multi-step reasoning. In this paper, we systematically study how to enhance the reasoning ability of LLMs for text-to-SQL parsing through chain-of-thought (CoT) style promptings including CoT prompting and Least-to-Most prompting. Our experiments demonstrate that iterative prompting as in Least-to-Most prompting may be unnecessary for text-to-SQL parsing and directly applying existing CoT style prompting methods leads to error propagation issues. By improving multi-step reasoning while avoiding much detailed information in the reasoning steps which may lead to error propagation, our new method outperforms existing ones by 2.4 point absolute gains on the Spider development set.

Abstract: In this work we try to apply Large Language Models (LLMs) to reframe the Question Answering task as Programming (QAaP). Due to the inherent dynamic nature of the real world, factual questions frequently involve a symbolic constraint: time, solving these questions necessitates not only extensive world knowledge, but also advanced reasoning ability to satisfy the temporal constraints. Despite the remarkable intelligence exhibited by LLMs in various NLP tasks, our experiments reveal that the aforementioned problems continue to pose a significant challenge to existing LLMs. To solve these time-sensitive factual questions, considering that modern LLMs possess superior ability in both natural language understanding and programming,we endeavor to leverage LLMs to represent diversely expressed text as well-structured code, and thereby grasp the desired knowledge along with the underlying symbolic constraint.

Abstract: Multilingual sequence-to-sequence models perform poorly with increased language coverage and fail to consistently generate text in the correct target language in few-shot settings. To address these challenges, we propose mmT5, a modular multilingual sequence-to-sequence model. mmT5 utilizes language-specific modules during pre-training, which disentangle language-specific information from language-agnostic information. We identify representation drift during fine-tuning as a key limitation of modular generative models and develop strategies that enable effective zero-shot transfer. Our model outperforms mT5 at the same parameter sizes by a large margin on representative natural language understanding and generation tasks in 40+ languages. Compared to mT5, mmT5 raises the rate of generating text in the correct language under zero-shot settings from 7% to 99%, thereby greatly alleviating the source language hallucination problem.

Abstract: Considerable advancements have been made to tackle the misrepresentation of information derived from reference articles in the domains of fact-checking and faithful summarization. However, an unaddressed aspect remains - the identification of social media posts that manipulate information within associated news articles. This task presents a significant challenge, primarily due to the prevalence of personal opinions in such posts. We present a novel task, identifying manipulation of news on social media, which aims to detect manipulation in social media posts and identify manipulated or inserted information. To study this task, we have proposed a data collection schema and curated a dataset called ManiTweet, consisting of 3.6K pairs of tweets and corresponding articles. Our analysis demonstrates that this task is highly challenging, with large language models (LLMs) yielding unsatisfactory performance. Additionally, we have developed a simple yet effective basic model that outperforms LLMs significantly on the ManiTweet dataset. Finally, we have conducted an exploratory analysis of human-written tweets, unveiling intriguing connections between manipulation and the domain and factuality of news articles, as well as revealing that manipulated sentences are more likely to encapsulate the main story or consequences of a news outlet.

Abstract: Multilingual machine translation has proven immensely useful for low-resource and zero-shot language pairs. However, language pairs in multilingual models sometimes see worse performance than in bilingual models, especially when translating in a one-to-many setting. To understand why, we examine the geometric differences in the representations from bilingual models versus those from one-to-many multilingual models. Specifically, we evaluate the isotropy of the representations, to measure how well they utilize the dimensions in their underlying vector space. Using the same evaluation data in both models, we find that multilingual model decoder representations tend to be less isotropic than bilingual model decoder representations. Additionally, we show that much of the anisotropy in multilingual decoder representations can be attributed to modeling language-specific information, therefore limiting remaining representational capacity.

Abstract: Scientific literature understanding tasks have gained significant attention due to their potential to accelerate scientific discovery. Pre-trained language models (LMs) have shown effectiveness in these tasks, especially when tuned via contrastive learning. However, jointly utilizing pre-training data across multiple heterogeneous tasks (e.g., extreme classification, citation prediction, and literature search) remains largely unexplored. To bridge this gap, we propose a multi-task contrastive learning framework, SciMult, with a focus on facilitating common knowledge sharing across different scientific literature understanding tasks while preventing task-specific skills from interfering with each other. To be specific, we explore two techniques -- task-aware specialization and instruction tuning. The former adopts a Mixture-of-Experts Transformer architecture with task-aware sub-layers; the latter prepends task-specific instructions to the input text so as to produce task-aware outputs. Extensive experiments on a comprehensive collection of benchmark datasets verify the effectiveness of our task-aware specialization strategy in various tasks, where we outperform state-of-the-art scientific LMs.

Abstract: Fine-tuning on instruction data has been widely validated as an effective practice for implementing chat language models like ChatGPT. Scaling the diversity and quality of such data, although straightforward, stands a great chance of leading to improved performance. This paper aims to improve the upper bound of open-source models further. We first provide a systematically designed, diverse, informative, large-scale dataset of instructional conversations, UltraChat, which does not involve human queries. Our objective is to capture the breadth of interactions that a human might have with an AI assistant and employs a comprehensive framework to generate multi-turn conversation iteratively. UltraChat contains 1.5 million high-quality multi-turn dialogues and covers a wide range of topics and instructions. Our statistical analysis of UltraChat reveals its superiority in various key metrics, including scale, average length, diversity, coherence, etc., solidifying its position as a leading open-source dataset. Building upon UltraChat, we fine-tune a LLaMA model to create a powerful conversational model, UltraLLaMA. Our evaluations indicate that UltraLLaMA consistently outperforms other open-source models, including Vicuna, the previously recognized state-of-the-art open-source model. The dataset and the model will be publicly released\footnote{\url{https://github.com/thunlp/UltraChat}}.

Abstract: Multilingual Large Language Models (LLMs) have recently shown great capability in various tasks, exhibiting state-of-the-art performance using few-shot or zero-shot prompting methods. While these models have been extensively studied in tasks where inputs are assumed to be in a single language, less attention has been paid to exploring their performance when inputs involve code-switching (CSW). In this paper, we provide an extensive empirical study of various multilingual LLMs and benchmark their performance in three tasks: sentiment analysis, machine translation, and word-level language identification. Our findings indicate that despite multilingual LLMs showing promising outcomes in certain tasks when using zero-/few-shot prompting, their performance still falls short on average when compared to smaller finetuned models. We argue that LLMs that are "multilingual" are not necessarily code-switching compatible and extensive future research is required to fully bridge this gap.

Abstract: Explainable multi-hop question answering (QA) not only predicts answers but also identifies rationales, i. e. subsets of input sentences used to derive the answers. This problem has been extensively studied under the supervised setting, where both answer and rationale annotations are given. Because rationale annotations are expensive to collect and not always available, recent efforts have been devoted to developing methods that do not rely on supervision for rationales. However, such methods have limited capacities in modeling interactions between sentences, let alone reasoning across multiple documents. This work proposes a principled, probabilistic approach for training explainable multi-hop QA systems without rationale supervision. Our approach performs multi-hop reasoning by explicitly modeling rationales as sets, enabling the model to capture interactions between documents and sentences within a document. Experimental results show that our approach is more accurate at selecting rationales than the previous methods, while maintaining similar accuracy in predicting answers.

Abstract: Recent studies have found that summaries generated by large language models (LLMs) are favored by human annotators over the original reference summaries in commonly used summarization datasets. Therefore, we investigate a new learning paradigm of text summarization models that considers the LLMs as the reference or the gold-standard oracle on commonly used summarization datasets such as the CNN/DailyMail dataset. To examine the standard practices that are aligned with the new learning setting, we propose a novel training method that is based on contrastive learning with LLMs as a summarization quality evaluator. For this reward-based training method, we investigate two different methods of utilizing LLMs for summary quality evaluation, namely GPTScore and GPTRank. Our experiments on the CNN/DailyMail dataset demonstrate that smaller summarization models trained by our proposed method can achieve performance equal to or surpass that of the reference LLMs, as evaluated by the LLMs themselves. This underscores the efficacy of our proposed paradigm in enhancing model performance over the standard maximum likelihood estimation (MLE) training method, and its efficiency since it only requires a small budget to access the LLMs. We release the training scripts, model outputs, and LLM-based evaluation results to facilitate future studies.

Abstract: Machine Translation (MT) has been widely used for cross-lingual classification, either by translating the test set into English and running inference with a monolingual model (translate-test), or translating the training set into the target languages and finetuning a multilingual model (translate-train). However, most research in the area focuses on the multilingual models rather than the MT component. We show that, by using a stronger MT system and mitigating the mismatch between training on original text and running inference on machine translated text, translate-test can do substantially better than previously assumed. The optimal approach, however, is highly task dependent, as we identify various sources of cross-lingual transfer gap that affect different tasks and approaches differently. Our work calls into question the dominance of multilingual models for cross-lingual classification, and prompts to pay more attention to MT-based baselines.

Abstract: The most meaningful connections between people are often fostered through expression of shared vulnerability and emotional experiences in personal narratives. We introduce a new task of identifying similarity in personal stories based on empathic resonance, i.e., the extent to which two people empathize with each others' experiences, as opposed to raw semantic or lexical similarity, as has predominantly been studied in NLP. Using insights from social psychology, we craft a framework that operationalizes empathic similarity in terms of three key features of stories: main events, emotional trajectories, and overall morals or takeaways. We create EmpathicStories, a dataset of 1,500 personal stories annotated with our empathic similarity features, and 2,000 pairs of stories annotated with empathic similarity scores. Using our dataset, we fine-tune a model to compute empathic similarity of story pairs, and show that this outperforms semantic similarity models on automated correlation and retrieval metrics. Through a user study with 150 participants, we also assess the effect our model has on retrieving stories that users empathize with, compared to naive semantic similarity-based retrieval, and find that participants empathized significantly more with stories retrieved by our model. Our work has strong implications for the use of empathy-aware models to foster human connection and empathy between people.

Abstract: While large language models (LLMs) are proficient at question-answering (QA), the dependencies between their answers and other "beliefs" they may have about the world are typically unstated, and may even be in conflict. Our goal is to uncover such dependencies and reduce inconsistencies among them, so that answers are supported by faithful, system-believed chains of reasoning drawn from a consistent network of beliefs. Our approach, which we call REFLEX, is to add a "rational", self-reflecting layer on top of the LLM. First, given a question, we construct a belief graph using a backward-chaining process to materialize relevant model "beliefs" (including beliefs about answer candidates) and the inferential relationships between them. Second, we identify and minimize contradictions in that graph using a formal constraint reasoner. We find that REFLEX significantly improves consistency (by 8%-11% absolute) without harming overall answer accuracy, resulting in answers supported by faithful chains of reasoning drawn from a more consistent belief system. This suggests a new style of system architecture, in which an LLM extended with a rational layer of self-reflection can repair latent inconsistencies within the LLM alone.

Abstract: Evaluating the factuality of long-form text generated by large language models (LMs) is non-trivial because (1) generations often contain a mixture of supported and unsupported pieces of information, making binary judgments of quality inadequate, and (2) human evaluation is time-consuming and costly. In this paper, we introduce FActScore (Factual precision in Atomicity Score), a new evaluation that breaks a generation into a series of atomic facts and computes the percentage of atomic facts supported by a reliable knowledge source. We conduct an extensive human evaluation to obtain FActScores of people biographies generated by several state-of-the-art commercial LMs -- InstructGPT, ChatGPT, and the retrieval-augmented PerplexityAI -- and report new analysis demonstrating the need for such a fine-grained score (e.g., ChatGPT only achieves 58%). Since human evaluation is costly, we also introduce an automated model that estimates FActScore, using retrieval and a strong language model, with less than a 2% error rate. Finally, we use this automated metric to evaluate 6,500 generations from a new set of 13 recent LMs that would have cost $26K if evaluated by humans, with various findings: GPT-4 and ChatGPT are more factual than public models, and Vicuna and Alpaca are some of the best public models.

Abstract: Semantics of a sentence is defined with much less ambiguity than semantics of a single word, and it should be better preserved by translation to another language. If multilingual sentence embeddings intend to represent sentence semantics, then the similarity between embeddings of any two sentences must be invariant with respect to translation. Based on this suggestion, we consider a simple linear cross-lingual mapping as a possible improvement of the multilingual embeddings. We also consider deviation from orthogonality conditions as a measure of deficiency of the embeddings.

Abstract: Large language models (LLMs) struggle on processing complicated observations in interactive decision making. To alleviate this issue, we propose a simple hierarchical prompting approach. Diverging from previous prompting approaches that always put the \emph{full} observation~(\eg a web page) to the prompt, we propose to first construct an action-aware observation which is more \emph{condensed} and \emph{relevant} with a dedicated \summ prompt. The \actor prompt then predicts the next action based on the summarized history. While our method has broad applicability, we particularly demonstrate its efficacy in the complex domain of web navigation where a full observation often contains redundant and irrelevant information. Our approach outperforms the previous state-of-the-art prompting mechanism with the same LLM by 6.2\% on task success rate, demonstrating its potential on interactive decision making tasks with long observation traces.

Abstract: Literature-Based Discovery (LBD) aims to discover new scientific knowledge by mining papers and generating hypotheses. Standard LBD is limited to predicting pairwise relations between discrete concepts (e.g., drug-disease links). LBD also ignores critical contexts like experimental settings (e.g., a specific patient population where a drug is evaluated) and background knowledge and motivations that human scientists consider (e.g., to find a drug candidate without specific side effects). We address these limitations with a novel formulation of contextualized-LBD (C-LBD): generating scientific hypotheses in natural language, while grounding them in a context that controls the hypothesis search space. We present a new modeling framework using retrieval of ``inspirations'' from a heterogeneous network of citations and knowledge graph relations, and create a new dataset derived from papers. In automated and human evaluations, our models improve over baselines, including powerful large language models (LLMs), but also reveal challenges on the road to building machines that generate new scientific knowledge.

Abstract: The ability to assist humans during a navigation task in a supportive role is crucial for intelligent agents. Such agents, equipped with environment knowledge and conversational abilities, can guide individuals through unfamiliar terrains by generating natural language responses to their inquiries, grounded in the visual information of their surroundings. However, these multimodal conversational navigation helpers are still underdeveloped. This paper proposes a new benchmark, Respond to Help (R2H), to build multimodal navigation helpers that can respond to help, based on existing dialog-based embodied datasets. R2H mainly includes two tasks: (1) Respond to Dialog History (RDH), which assesses the helper agent's ability to generate informative responses based on a given dialog history, and (2) Respond during Interaction (RdI), which evaluates the helper agent's ability to maintain effective and consistent cooperation with a task performer agent during navigation in real-time. Furthermore, we propose a novel task-oriented multimodal response generation model that can see and respond, named SeeRee, as the navigation helper to guide the task performer in embodied tasks. Through both automatic and human evaluations, we show that SeeRee produces more effective and informative responses than baseline methods in assisting the task performer with different navigation tasks. Project website: https://sites.google.com/view/respond2help/home.

Abstract: Knowing the language of an input text/audio is a necessary first step for using almost every natural language processing (NLP) tool such as taggers, parsers, or translation systems. Language identification is a well-studied problem, sometimes even considered solved; in reality, most of the world's 7000 languages are not supported by current systems. This lack of representation affects large-scale data mining efforts and further exacerbates data shortage for low-resource languages. We take a step towards tackling the data bottleneck by compiling a corpus of over 50K parallel children's stories in 350+ languages and dialects, and the computation bottleneck by building lightweight hierarchical models for language identification. Our data can serve as benchmark data for language identification of short texts and for understudied translation directions such as those between Indian or African languages. Our proposed method, Hierarchical LIMIT, uses limited computation to expand coverage into excluded languages while maintaining prediction quality.

Abstract: The remarkable advancements in large language models (LLMs) have significantly enhanced the performance in few-shot learning settings. By using only a small number of labeled examples, referred to as demonstrations, LLMs can effectively grasp the task at hand through in-context learning. However, the process of selecting appropriate demonstrations has received limited attention in prior work. This paper addresses the issue of identifying the most informative demonstrations for few-shot learning by approaching it as a pool-based Active Learning (AL) problem over a single iteration. Our objective is to investigate how AL algorithms can serve as effective demonstration selection methods for in-context learning. We compare various standard AL algorithms based on uncertainty, diversity, and similarity, and consistently observe that the latter outperforms all other methods, including random sampling. Notably, uncertainty sampling, despite its success in conventional supervised learning scenarios, performs poorly in this context. Our extensive experimentation involving a diverse range of GPT and OPT models across $24$ classification and multi-choice tasks, coupled with thorough analysis, unambiguously demonstrates that in-context example selection through AL prioritizes high-quality examples that exhibit low uncertainty and bear similarity to the test examples.

Abstract: Large language models (LLMs) have achieved widespread success on a variety of in-context few-shot tasks, but this success is typically evaluated via correctness rather than consistency. We argue that self-consistency is an important criteria for valid multi-step reasoning and propose two types of self-consistency that are particularly important for multi-step logic -- hypothetical consistency (the ability for a model to predict what its output would be in a hypothetical other context) and compositional consistency (consistency of a model's outputs for a compositional task even when an intermediate step is replaced with the model's output for that step). We demonstrate that four sizes of the GPT-3 model exhibit poor consistency rates across both types of consistency on four different tasks (Wikipedia, DailyDialog, arithmetic, and GeoQuery).

Abstract: We introduce and demonstrate how to effectively train multilingual machine translation models with pixel representations. We experiment with two different data settings with a variety of language and script coverage, and show performance competitive with subword embeddings. We analyze various properties of pixel representations to better understand where they provide potential benefits and the impact of different scripts and data representations. We observe that these properties not only enable seamless cross-lingual transfer to unseen scripts, but make pixel representations more data-efficient than alternatives such as vocabulary expansion. We hope this work contributes to more extensible multilingual models for all languages and scripts.

Abstract: Recent work in vision-and-language pretraining has investigated supervised signals from object detection data to learn better, fine-grained multimodal representations. In this work, we take a step further and explore how we add supervision from small-scale visual relation data. In particular, we propose two pretraining approaches to contextualise visual entities in a multimodal setup. With verbalised scene graphs, we transform visual relation triplets into structured captions, and treat them as additional views of images. With masked relation prediction, we further encourage relating entities from visually masked contexts. When applied to strong baselines pretrained on large amounts of Web data, zero-shot evaluations on both coarse-grained and fine-grained tasks show the efficacy of our methods in learning multimodal representations from weakly-supervised relations data.

Abstract: The field of automatic evaluation of text generation made tremendous progress in the last few years. In particular, since the advent of neural metrics, like COMET, BLEURT, and SEScore2, the newest generation of metrics show a high correlation with human judgment. Unfortunately, quality scores generated with neural metrics are not interpretable, and it is unclear which part of the generation output is criticized by the metrics. To address this limitation, we present INSTRUCTSCORE, an open-source, explainable evaluation metric for text generation. By harnessing both explicit human instruction and the implicit knowledge of GPT4, we fine-tune a LLAMA model to create an evaluative metric that can produce a diagnostic report aligned with human judgment. We evaluate INSTRUCTSCORE on the WMT22 Zh-En translation task, where our 7B model surpasses other LLM-based baselines, including those based on 175B GPT3. Impressively, our INSTRUCTSCORE, even without direct supervision from human-rated data, achieves performance levels on par with state-of-the-art metrics like COMET22, which was fine-tuned on human ratings.

Abstract: Large Language Models (LLMs) play a powerful \textit{Reader} of the \textit{Retrieve-then-Read} pipeline, making great progress in knowledge-based open-domain tasks. This work introduces a new framework, \textit{Rewrite-Retrieve-Read} that improves the retrieval-augmented method from the perspective of the query rewriting. Prior studies mostly contribute to adapt the retriever or stimulate the reader. Different from them, our approach pay attention of the query adaptation. Because the original query can not be always optimal to retrieve for the LLM, especially in the real world.(1) We first prompt an LLM to rewrite the queries, then conduct retrieval-augmented reading. (2) We further apply a small language model as a trainable rewriter, which rewrite the search query to cater to the frozen retriever and the LLM reader. To fine-tune the rewriter, we first use a pseudo data to conduct supervised warm-up training. Then the \textit{Retrieve-then-Read} pipeline is modeled as a reinforcement learning context. The rewriter is further trained as a policy model by maximize the reward of the pipeline performance. Evaluation is performed on two downstream tasks, open-domain QA and multiple choice. Our framework is proved effective and scalable.

Abstract: This paper aims to explore the potential of leveraging Large Language Models (LLMs) for data augmentation in crosslingual commonsense reasoning datasets, where the available training data is extremely limited. To achieve this, we employ several LLMs including Dolly-v2, StableVicuna, ChatGPT, and GPT-4 to augment three datasets: XCOPA, XWinograd, and XStoryCloze. Subsequently, we assess the effectiveness of fine-tuning smaller crosslingual models, mBERT and XLMR, using the synthesised data. We compare the performance of training with data generated in English and target languages, as well as translating the English-generated data into the target languages. Our experiments reveal the overall advantages of incorporating data generated by LLMs. Training on synthetic data generated by GPT-4, whether English or multilingual, improves performance consistently compared to the baseline. Other models also exhibit an overall increase in performance, however, their effectiveness decreases in some settings. We also ask native speakers to evaluate the naturalness and logical soundness of the generated examples for different languages. Human evaluation reveals that LLMs like ChatGPT and GPT-4 excel at generating natural text in most languages, except a few such as Tamil. Moreover, ChatGPT trails behind in generating plausible alternatives in comparison to the original dataset, while GPT-4 demonstrates competitive logic consistency in the synthesised data.

Abstract: We evaluate how well LLMs understand African American Language (AAL) in comparison to their performance on White Mainstream English (WME), the encouraged "standard" form of English taught in American classrooms. We measure LLM performance using automatic metrics and human judgments for two tasks: a counterpart generation task, where a model generates AAL (or WME) given WME (or AAL), and a masked span prediction (MSP) task, where models predict a phrase that was removed from their input. Our contributions include: (1) evaluation of six pre-trained, large language models on the two language generation tasks; (2) a novel dataset of AAL text from multiple contexts (social media, hip-hop lyrics, focus groups, and linguistic interviews) with human-annotated counterparts in WME; and (3) documentation of model performance gaps that suggest bias and identification of trends in lack of understanding of AAL features.

Abstract: Despite recent advances in Large Language Models (LLMs), users still cannot trust the information provided in their responses. LLMs cannot speak accurately about events that occurred after their training, which are often topics of great interest to users, and, as we show in this paper, they are highly prone to hallucination when talking about less popular (tail) topics. This paper presents WikiChat, a few-shot LLM-based chatbot that is grounded with live information from Wikipedia. Through many iterations of experimentation, we have crafte a pipeline based on information retrieval that (1) uses LLMs to suggest interesting and relevant facts that are individually verified against Wikipedia, (2) retrieves additional up-to-date information, and (3) composes coherent and engaging time-aware responses. We propose a novel hybrid human-and-LLM evaluation methodology to analyze the factuality and conversationality of LLM-based chatbots. We focus on evaluating important but previously neglected issues such as conversing about recent and tail topics. We evaluate WikiChat against strong fine-tuned and LLM-based baselines across a diverse set of conversation topics. We find that WikiChat outperforms all baselines in terms of the factual accuracy of its claims, by up to 12.1%, 28.3% and 32.7% on head, recent and tail topics, while matching GPT-3.5 in terms of providing natural, relevant, non-repetitive and informational responses.

Abstract: Extracting structured and grounded fact triples from raw text is a fundamental task in Information Extraction (IE). Existing IE datasets are typically collected from Wikipedia articles, using hyperlinks to link entities to the Wikidata knowledge base. However, models trained only on Wikipedia have limitations when applied to web domains, which often contain noisy text or text that does not have any factual information. We present WebIE, the first large-scale, entity-linked closed IE dataset consisting of 1.6M sentences automatically collected from the English Common Crawl corpus. WebIE also includes negative examples, i.e. sentences without fact triples, to better reflect the data on the web. We annotate ~25K triples from WebIE through crowdsourcing and introduce mWebIE, a translation of the annotated set in four other languages: French, Spanish, Portuguese, and Hindi. We evaluate the in-domain, out-of-domain, and zero-shot cross-lingual performance of generative IE models and find models trained on WebIE show better generalisability. We also propose three training strategies that use entity linking as an auxiliary task. Our experiments show that adding Entity-Linking objectives improves the faithfulness of our generative IE models.

Abstract: An abundance of datasets exist for training and evaluating models on the task of summary generation.However, these datasets are often derived heuristically, and lack sufficient annotations to support research into all aspects of summarization, such as evidence extraction and controllable summarization. We introduce a benchmark comprising 8 tasks that require multi-dimensional understanding of summarization, e.g., surfacing evidence for a summary, assessing its correctness, and gauging its relevance to different topics. We compare various methods on this benchmark and discover that on multiple tasks, moderately-sized fine-tuned models consistently outperform much larger few-shot prompted language models. For factuality related tasks, we also evaluate existing heuristics to create training data and find that training on them performs worse than training on $20\times$ less human-labeled data. Our benchmark consists of data from 6 different domains, allowing us to study cross-domain performance of trained models. We find that for some tasks, the amount of training data matters more than the domain where it comes from, while for other tasks training specifically on data from the target domain, even if limited, is more beneficial. Our work fulfills the need for a well-annotated summarization benchmark with diverse tasks, and provides useful insights about the impact of the quality, size and domain of training data.

Abstract: Learning high quality sentence embeddings from dialogues has drawn increasing attentions as it is essential to solve a variety of dialogue-oriented tasks with low annotation cost. However, directly annotating and gathering utterance relationships in conversations are difficult, while token-level annotations, \eg, entities, slots and templates, are much easier to obtain. General sentence embedding methods are usually sentence-level self-supervised frameworks and cannot utilize token-level extra knowledge. In this paper, we introduce Template-aware Dialogue Sentence Embedding (TaDSE), a novel augmentation method that utilizes template information to effectively learn utterance representation via self-supervised contrastive learning framework. TaDSE augments each sentence with its corresponding template and then conducts pairwise contrastive learning over both sentence and template. We further enhance the effect with a synthetically augmented dataset that enhances utterance-template relation, in which entity detection (slot-filling) is a preliminary step. We evaluate TaDSE performance on five downstream benchmark datasets. The experiment results show that TaDSE achieves significant improvements over previous SOTA methods, along with a consistent Intent Classification task performance improvement margin. We further introduce a novel analytic instrument of Semantic Compression method, for which we discover a correlation with uniformity and alignment. Our code will be released soon.

Abstract: People primarily consult tables to conduct data analysis or answer specific questions. Text generation systems that can provide accurate table summaries tailored to users' information needs can facilitate more efficient access to relevant data insights. However, existing table-to-text generation studies primarily focus on converting tabular data into coherent statements, rather than addressing information-seeking purposes. In this paper, we define a new query-focused table summarization task, where text generation models have to perform human-like reasoning and analysis over the given table to generate a tailored summary, and we introduce a new benchmark named QTSumm for this task. QTSumm consists of 5,625 human-annotated query-summary pairs over 2,437 tables on diverse topics. Moreover, we investigate state-of-the-art models (i.e., text generation, table-to-text generation, and large language models) on the QTSumm dataset. Experimental results and manual analysis reveal that our benchmark presents significant challenges in table-to-text generation for future research.

Abstract: Debiasing methods that seek to mitigate the tendency of Language Models (LMs) to occasionally output toxic or inappropriate text have recently gained traction. In this paper, we propose a standardized protocol which distinguishes methods that yield not only desirable results, but are also consistent with their mechanisms and specifications. For example, we ask, given a debiasing method that is developed to reduce toxicity in LMs, if the definition of toxicity used by the debiasing method is reversed, would the debiasing results also be reversed? We used such considerations to devise three criteria for our new protocol: Specification Polarity, Specification Importance, and Domain Transferability. As a case study, we apply our protocol to a popular debiasing method, Self-Debiasing, and compare it to one we propose, called Instructive Debiasing, and demonstrate that consistency is as important an aspect to debiasing viability as is simply a desirable result. We show that our protocol provides essential insights into the generalizability and interpretability of debiasing methods that may otherwise go overlooked.

Abstract: Instruction-tuned Large Language Models (LLMs) have exhibited impressive language understanding and the capacity to generate responses that follow specific instructions. However, due to the computational demands associated with training these models, their applications often rely on zero-shot settings. In this paper, we evaluate the zero-shot performance of two publicly accessible LLMs, ChatGPT and OpenAssistant, in the context of Computational Social Science classification tasks, while also investigating the effects of various prompting strategies. Our experiment considers the impact of prompt complexity, including the effect of incorporating label definitions into the prompt, using synonyms for label names, and the influence of integrating past memories during the foundation model training. The findings indicate that in a zero-shot setting, the current LLMs are unable to match the performance of smaller, fine-tuned baseline transformer models (such as BERT). Additionally, we find that different prompting strategies can significantly affect classification accuracy, with variations in accuracy and F1 scores exceeding 10%.

Abstract: Large Language Models (LLMs) have demonstrated significant progress in utilizing external APIs as tools for various tasks. However, their tool-using ability is limited by the availability of suitable APIs and the instability of implicit reasoning, particularly when simultaneously engaging in reasoning about plans and actual calculations. To address these limitations, we propose CREATOR, a novel framework that empowers LLMs to create their own tools through documentation and code realization. CREATOR disentangles the LLM's ability into two distinct phases: abstract tool creation and concrete decision execution, which results in improved LLM performance. We evaluate CREATOR on two established benchmarks: MATH, which consists of challenging math competition problems, and TabMWP, which includes diverse tabular contents for problem-solving. Remarkably, CREATOR significantly outperforms existing chain-of-thought (CoT), program-of-thought (PoT), and tool-using baselines on these two benchmarks. Additionally, we present a new dataset, Creation Challenge, comprising 2K diverse questions, to highlight the necessity and benefits of LLMs' tool creation ability in effectively addressing these problems. Furthermore, our research reveals that leveraging LLMs as tool creators facilitates knowledge transfer, and LLMs exhibit varying levels of tool creation abilities, enabling them to flexibly tackle diverse situations. Our study represents a promising avenue for maximizing the potential of LLMs and advancing toward truly intelligent and adaptable AI systems.

Abstract: We propose ConGraT(Contrastive Graph-Text pretraining), a general, self-supervised method for jointly learning separate representations of texts and nodes in a parent (or ``supervening'') graph, where each text is associated with one of the nodes. Datasets fitting this paradigm are common, from social media (users and posts), to citation networks over articles, to link graphs over web pages. We expand on prior work by providing a general, self-supervised, joint pretraining method, one which does not depend on particular dataset structure or a specific task. Our method uses two separate encoders for graph nodes and texts, which are trained to align their representations within a common latent space. Training uses a batch-wise contrastive learning objective inspired by prior work on joint text and image encoding. As graphs are more structured objects than images, we also extend the training objective to incorporate information about node similarity and plausible next guesses in matching nodes and texts. Experiments on various datasets reveal that ConGraT outperforms strong baselines on various downstream tasks, including node and text category classification and link prediction. Code and certain datasets are available at https://github.com/wwbrannon/congrat.

Abstract: Large language models (LLMs) have significantly advanced the field of natural language processing (NLP) through their extensive parameters and comprehensive data utilization. However, existing LLMs lack a dedicated memory unit, limiting their ability to explicitly store and retrieve knowledge for various tasks. In this paper, we propose RET-LLM a novel framework that equips LLMs with a general write-read memory unit, allowing them to extract, store, and recall knowledge from the text as needed for task performance. Inspired by Davidsonian semantics theory, we extract and save knowledge in the form of triplets. The memory unit is designed to be scalable, aggregatable, updatable, and interpretable. Through qualitative evaluations, we demonstrate the superiority of our proposed framework over baseline approaches in question answering tasks. Moreover, our framework exhibits robust performance in handling temporal-based question answering tasks, showcasing its ability to effectively manage time-dependent information.

Abstract: Although large language models (LLMs) have achieved excellent performance in a variety of evaluation benchmarks, they still struggle in complex reasoning tasks which require specific knowledge and multi-hop reasoning. To improve the reasoning abilities, we propose \textbf{ChatCoT}, a tool-augmented chain-of-thought reasoning framework for chat-based LLMs. In ChatCoT, we model the chain-of-thought~(CoT) reasoning as multi-turn conversations, to utilize tools in a more natural way through chatting. At each turn, LLMs can either interact with tools or perform the reasoning. %framework to decompose reasoning and manipulate tools for complex reasoning tasks via multi-round dialogue. %Our idea is to decompose the complex problems into several sub-problems and solve these sub-problems with the help of external tools. Our approach can effectively leverage the multi-turn conversation ability of chat-based LLMs, and integrate the thought chain following and tools manipulation in a unified way. Specially, we initialize the early turns of the conversation by the tools, tasks and reasoning format, and propose an iterative \emph{tool-augmented reasoning} step to perform step-by-step tool-augmented reasoning. The experiment results on two complex reasoning datasets (MATH and HotpotQA) have shown the effectiveness of ChatCoT on complex reasoning tasks, achieving a 6.8\% relative improvement over the state-of-the-art baseline. %and ChatCoT can achieve performance on Math. Our code and data are available at: \url{https://github.com/RUCAIBOX/ChatCoT}.

Abstract: Kendall's tau is frequently used to meta-evaluate how well machine translation (MT) evaluation metrics score individual translations. Its focus on pairwise score comparisons is intuitive but raises the question of how ties should be handled, a gray area that has motivated different variants in the literature. We demonstrate that, in settings like modern MT meta-evaluation, existing variants have weaknesses arising from their handling of ties, and in some situations can even be gamed. We propose a novel variant that gives metrics credit for correctly predicting ties, as well as an optimization procedure that automatically introduces ties into metric scores, enabling fair comparison between metrics that do and do not predict ties. We argue and provide experimental evidence that these modifications lead to fairer Kendall-based assessments of metric performance.

Abstract: Large language models (LLMs) have demonstrated remarkable capabilities in language generation, understanding, and few-shot learning in recent years. An extensive body of work has explored how their performance may be further improved through the tools of prompting, ranging from verification, self-consistency, or intermediate scratchpads. In this paper, we present a complementary approach to improve language responses where multiple language model instances propose and debate their individual responses and reasoning processes over multiple rounds to arrive at a common final answer. Our findings indicate that this approach significantly enhances mathematical and strategic reasoning across a number of tasks. We also demonstrate that our approach improves the factual validity of generated content, reducing fallacious answers and hallucinations that contemporary models are prone to. Our approach may be directly applied to existing black-box models and uses identical procedure and prompts for all tasks we investigate. Overall, our findings suggest that such "society of minds" approach has the potential to significantly advance the capabilities of LLMs and pave the way for further breakthroughs in language generation and understanding.

Abstract: Empowering language is important in many real-world contexts, from education to workplace dynamics to healthcare. Though language technologies are growing more prevalent in these contexts, empowerment has not been studied in NLP, and moreover, it is inherently challenging to operationalize because of its subtle, implicit nature. This work presents the first computational exploration of empowering language. We first define empowerment detection as a new task, grounding it in linguistic and social psychology literature. We then crowdsource a novel dataset of Reddit posts labeled for empowerment, reasons why these posts are empowering to readers, and the social relationships between posters and readers. Our preliminary analyses show that this dataset, which we call TalkUp, can be used to train language models that capture empowering and disempowering language. More broadly, as it is rich with the ambiguities and diverse interpretations of real-world language, TalkUp provides an avenue to explore implication, presuppositions, and how social context influences the meaning of language.

Abstract: Instruction tuning has emerged to enhance the capabilities of large language models (LLMs) in providing appropriate outputs based on input instructions. However, existing methods for collecting instruction-tuning data suffer from limitations in scalability and affordability. In this paper, we propose Dynosaur, a dynamic growth paradigm for instruction-tuning data curation. Built upon the metadata of existing NLP datasets, we generate multiple task instructions applicable to various NLP datasets and determine the relevant data fields for constructing instruction-tuning data with LLMs. Dynosaur offers several advantages: 1) lower generation costs (less than $12 for generating 800K instruction-tuning data), 2) good quality of instruction-tuning data (better performance than Alpaca and Instruction GPT-4 on Super-NI with comparable data sizes), and 3) the ability to grow dynamically by incorporating new datasets from Huggingface Datasets Platform. We further investigate continual learning as an approach to learning with the ever-growing instruction-tuning dataset. We demonstrate that replay methods not only help mitigate forgetting issues but help generalize to unseen tasks better. As a novel continual learning scenario for instruction tuning, selecting tasks based on instruction representations can be an effective replaying strategy. Code and data are released at \url{https://github.com/WadeYin9712/Dynosaur}.

Abstract: Traditional neural machine translation (NMT) systems often fail to translate sentences that contain culturally specific information. Most previous NMT methods have incorporated external cultural knowledge during training, which requires fine-tuning on low-frequency items specific to the culture. Recent in-context learning utilizes lightweight prompts to guide large language models (LLMs) to perform machine translation, however, whether such an approach works in terms of injecting culture awareness into machine translation remains unclear. To this end, we introduce a new data curation pipeline to construct a culturally relevant parallel corpus, enriched with annotations of cultural-specific entities. Additionally, we design simple but effective prompting strategies to assist this LLM-based translation. Extensive experiments show that our approaches can largely help incorporate cultural knowledge into LLM-based machine translation, outperforming traditional NMT systems in translating cultural-specific sentences.

Abstract: AI systems have shown impressive performance at answering questions by retrieving relevant context. However, with the increasingly large models, it is impossible and often undesirable to constrain models' knowledge or reasoning to only the retrieved context. This leads to a mismatch between the information that these models access to derive the answer and the information available to the user consuming the AI predictions to assess the AI predicted answer. In this work, we study how users interact with AI systems in absence of sufficient information to assess AI predictions. Further, we ask the question of whether adding the requisite background alleviates the concerns around over-reliance in AI predictions. Our study reveals that users rely on AI predictions even in the absence of sufficient information needed to assess its correctness. Providing the relevant background, however, helps users catch AI errors better, reducing over-reliance on incorrect AI predictions. On the flip side, background information also increases users' confidence in their correct as well as incorrect judgments. Contrary to common expectation, aiding a user's perusal of the context and the background through highlights is not helpful in alleviating the issue of over-confidence stemming from availability of more information. Our work aims to highlight the gap between how NLP developers perceive informational need in human-AI interaction and the actual human interaction with the information available to them.

Abstract: Trustworthy answer content is abundant in many high-resource languages and is instantly accessible through question answering systems, yet this content can be hard to access for those that do not speak these languages. The leap forward in cross-lingual modeling quality offered by generative language models offers much promise, yet their raw generations often fall short in factuality. To improve trustworthiness in these systems, a promising direction is to attribute the answer to a retrieved source, possibly in a content-rich language different from the query. Our work is the first to study attribution for cross-lingual question answering. First, we collect data in 5 languages to assess the attribution level of a state-of-the-art cross-lingual QA system. To our surprise, we find that a substantial portion of the answers is not attributable to any retrieved passages (up to 50% of answers exactly matching a gold reference) despite the system being able to attend directly to the retrieved text. Second, to address this poor attribution level, we experiment with a wide range of attribution detection techniques. We find that Natural Language Inference models and PaLM 2 fine-tuned on a very small amount of attribution data can accurately detect attribution. Based on these models, we improve the attribution level of a cross-lingual question-answering system. Overall, we show that current academic generative cross-lingual QA systems have substantial shortcomings in attribution and we build tooling to mitigate these issues.

Abstract: Chain-of-Thought and Program-Aided Language Models represent two distinct reasoning methods, each with its own strengths and weaknesses. We demonstrate that it is possible to combine the best of both worlds by using different models for different problems, employing a large language model (LLM) to perform model selection. Through a theoretical analysis, we discover that the performance improvement is determined by the differences between the combined methods and the success rate of choosing the correct model. On eight reasoning datasets, our proposed approach shows significant improvements. Furthermore, we achieve new state-of-the-art results on GSM8K and SVAMP with accuracies of 96.5% and 93.7%, respectively. Our code is publicly available at https://github.com/XuZhao0/Model-Selection-Reasoning.

Abstract: In this paper, we explore the question of whether language models (LLMs) can support cost-efficient information extraction from complex tables. We introduce schema-driven information extraction, a new task that uses LLMs to transform tabular data into structured records following a human-authored schema. To assess various LLM's capabilities on this task, we develop a benchmark composed of tables from three diverse domains: machine learning papers, chemistry tables, and webpages. Accompanying the benchmark, we present InstrucTE, a table extraction method based on instruction-tuned LLMs. This method necessitates only a human-constructed extraction schema, and incorporates an error-recovery strategy. Notably, InstrucTE demonstrates competitive performance without task-specific labels, achieving an F1 score ranging from 72.3 to 95.7. Moreover, we validate the feasibility of distilling more compact table extraction models to minimize extraction costs and reduce API reliance. This study paves the way for the future development of instruction-following models for cost-efficient table extraction.

Abstract: Internet links enable users to deepen their understanding of a topic by providing convenient access to related information. However, the majority of links are unanchored -- they link to a target webpage as a whole, and readers may expend considerable effort localizing the specific parts of the target webpage that enrich their understanding of the link's source context. To help readers effectively find information in linked webpages, we introduce the task of anchor prediction, where the goal is to identify the specific part of the linked target webpage that is most related to the source linking context. We release the AuthorAnchors dataset, a collection of 34K naturally-occurring anchored links, which reflect relevance judgments by the authors of the source article. To model reader relevance judgments, we annotate and release ReaderAnchors, an evaluation set of anchors that readers find useful. Our analysis shows that effective anchor prediction often requires jointly reasoning over lengthy source and target webpages to determine their implicit relations and identify parts of the target webpage that are related but not redundant. We benchmark a performant T5-based ranking approach to establish baseline performance on the task, finding ample room for improvement.

Abstract: While there has been significant development of models for Plain Language Summarization (PLS), evaluation remains a challenge. This is in part because PLS involves multiple, interrelated language transformations (e.g., adding background explanations, removing specialized terminology). No metrics are explicitly engineered for PLS, and the suitability of other text generation evaluation metrics remains unclear. To address these concerns, our study presents a granular meta-evaluation testbed, APPLS, designed to evaluate existing metrics for PLS. Drawing on insights from previous research, we define controlled perturbations for our testbed along four criteria that a metric of plain language should capture: informativeness, simplification, coherence, and faithfulness. Our analysis of metrics using this testbed reveals that current metrics fail to capture simplification, signaling a crucial gap. In response, we introduce POMME, a novel metric designed to assess text simplification in PLS. We demonstrate its correlation with simplification perturbations and validate across a variety of datasets. Our research contributes the first meta-evaluation testbed for PLS and a comprehensive evaluation of existing metrics, offering insights with relevance to other text generation tasks.