Computation and Language (cs.CL)

Abstract: In-context learning (ICL) for large language models has proven to be a powerful approach for many natural language processing tasks. However, determining the best method to select examples for ICL is nontrivial as the results can vary greatly depending on the quality, quantity, and order of examples used. In this paper, we conduct a case study on text simplification (TS) to investigate how to select the best and most robust examples for ICL. We propose Metric-Based in-context Learning (MBL) method that utilizes commonly used TS metrics such as SARI, compression ratio, and BERT-Precision for selection. Through an extensive set of experiments with various-sized GPT models on standard TS benchmarks such as TurkCorpus and ASSET, we show that examples selected by the top SARI scores perform the best on larger models such as GPT-175B, while the compression ratio generally performs better on smaller models such as GPT-13B and GPT-6.7B. Furthermore, we demonstrate that MBL is generally robust to example orderings and out-of-domain test sets, and outperforms strong baselines and state-of-the-art finetuned language models. Finally, we show that the behaviour of large GPT models can be implicitly controlled by the chosen metric. Our research provides a new framework for selecting examples in ICL, and demonstrates its effectiveness in text simplification tasks, breaking new ground for more accurate and efficient NLG systems.

Abstract: This paper addresses the classification of Arabic text data in the field of Natural Language Processing (NLP), with a particular focus on Natural Language Inference (NLI) and Contradiction Detection (CD). Arabic is considered a resource-poor language, meaning that there are few data sets available, which leads to limited availability of NLP methods. To overcome this limitation, we create a dedicated data set from publicly available resources. Subsequently, transformer-based machine learning models are being trained and evaluated. We find that a language-specific model (AraBERT) performs competitively with state-of-the-art multilingual approaches, when we apply linguistically informed pre-training methods such as Named Entity Recognition (NER). To our knowledge, this is the first large-scale evaluation for this task in Arabic, as well as the first application of multi-task pre-training in this context.

Abstract: Large language models (LLMs) have been widely employed for graph-to-text generation tasks. However, the process of finetuning LLMs requires significant training resources and annotation work. In this paper, we explore the capability of generative models to generate descriptive text from graph data in a zero-shot setting. Specifically, we evaluate GPT-3 and ChatGPT on two graph-to-text datasets and compare their performance with that of finetuned LLM models such as T5 and BART. Our results demonstrate that generative models are capable of generating fluent and coherent text, achieving BLEU scores of 10.57 and 11.08 for the AGENDA and WebNLG datasets, respectively. However, our error analysis reveals that generative models still struggle with understanding the semantic relations between entities, and they also tend to generate text with hallucinations or irrelevant information. As a part of error analysis, we utilize BERT to detect machine-generated text and achieve high macro-F1 scores. We have made the text generated by generative models publicly available.

Abstract: Whisper is one of the recent state-of-the-art multilingual speech recognition and translation models, however, it is not designed for real time transcription. In this paper, we build on top of Whisper and create Whisper-Streaming, an implementation of real-time speech transcription and translation of Whisper-like models. Whisper-Streaming uses local agreement policy with self-adaptive latency to enable streaming transcription. We show that Whisper-Streaming achieves high quality and 3.3 seconds latency on unsegmented long-form speech transcription test set, and we demonstrate its robustness and practical usability as a component in live transcription service at a multilingual conference.

Abstract: This paper presents a series of approaches aimed at enhancing the performance of Aspect-Based Sentiment Analysis (ABSA) by utilizing extracted semantic information from a Semantic Role Labeling (SRL) model. We propose a novel end-to-end Semantic Role Labeling model that effectively captures most of the structured semantic information within the Transformer hidden state. We believe that this end-to-end model is well-suited for our newly proposed models that incorporate semantic information. We evaluate the proposed models in two languages, English and Czech, employing ELECTRA-small models. Our combined models improve ABSA performance in both languages. Moreover, we achieved new state-of-the-art results on the Czech ABSA.

Abstract: In recent years, many NLP studies have focused solely on performance improvement. In this work, we focus on the linguistic and scientific aspects of NLP. We use the task of generating referring expressions in context (REG-in-context) as a case study and start our analysis from GREC, a comprehensive set of shared tasks in English that addressed this topic over a decade ago. We ask what the performance of models would be if we assessed them (1) on more realistic datasets, and (2) using more advanced methods. We test the models using different evaluation metrics and feature selection experiments. We conclude that GREC can no longer be regarded as offering a reliable assessment of models' ability to mimic human reference production, because the results are highly impacted by the choice of corpus and evaluation metrics. Our results also suggest that pre-trained language models are less dependent on the choice of corpus than classic Machine Learning models, and therefore make more robust class predictions.

Abstract: Paraphrasing is the task of expressing an essential idea or meaning in different words. But how different should the words be in order to be considered an acceptable paraphrase? And can we exclusively use automated metrics to evaluate the quality of a paraphrase? We attempt to answer these questions by conducting experiments on a German data set and performing automatic and expert linguistic evaluation.

Abstract: In this paper, we present the first application of Native Language Identification (NLI) for the Turkish language. NLI involves predicting the writer's first language by analysing their writing in different languages. While most NLI research has focused on English, our study extends its scope to Turkish. We used the recently constructed Turkish Learner Corpus and employed a combination of three syntactic features (CFG production rules, part-of-speech n-grams and function words) with L2 texts to demonstrate their effectiveness in this task.

Abstract: Archives play a crucial role in preserving information and knowledge, and the exponential growth of such data necessitates efficient and automated tools for managing and utilizing archive information resources. Archival applications involve managing massive data that are challenging to process and analyze. Although LLMs have made remarkable progress in diverse domains, there are no publicly available archives tailored LLM. Addressing this gap, we introduce ArcGPT, to our knowledge, the first general-purpose LLM tailored to the archival field. To enhance model performance on real-world archival tasks, ArcGPT has been pre-trained on massive and extensive archival domain data. Alongside ArcGPT, we release AMBLE, a benchmark comprising four real-world archival tasks. Evaluation on AMBLE shows that ArcGPT outperforms existing state-of-the-art models, marking a substantial step forward in effective archival data management. Ultimately, ArcGPT aims to better serve the archival community, aiding archivists in their crucial role of preserving and harnessing our collective information and knowledge.

Abstract: In-context learning, which offers substantial advantages over fine-tuning, is predominantly observed in decoder-only models, while encoder-decoder (i.e., seq2seq) models excel in methods that rely on weight updates. Recently, a few studies have demonstrated the feasibility of few-shot learning with seq2seq models; however, this has been limited to tasks that align well with the seq2seq architecture, such as summarization and translation. Inspired by these initial studies, we provide a first-ever extensive experiment comparing the in-context few-shot learning capabilities of decoder-only and encoder-decoder models on a broad range of tasks. Furthermore, we propose two methods to more effectively elicit in-context learning ability in seq2seq models: objective-aligned prompting and a fusion-based approach. Remarkably, our approach outperforms a decoder-only model that is six times larger and exhibits significant performance improvements compared to conventional seq2seq models across a variety of settings. We posit that, with the right configuration and prompt design, seq2seq models can be highly effective few-shot learners for a wide spectrum of applications.

Abstract: The Entity Set Expansion (ESE) task aims to expand a handful of seed entities with new entities belonging to the same semantic class. Conventional ESE methods are based on mono-modality (i.e., literal modality), which struggle to deal with complex entities in the real world such as: (1) Negative entities with fine-grained semantic differences. (2) Synonymous entities. (3) Polysemous entities. (4) Long-tailed entities. These challenges prompt us to propose Multi-modal Entity Set Expansion (MESE), where models integrate information from multiple modalities to represent entities. Intuitively, the benefits of multi-modal information for ESE are threefold: (1) Different modalities can provide complementary information. (2) Multi-modal information provides a unified signal via common visual properties for the same semantic class or entity. (3) Multi-modal information offers robust alignment signal for synonymous entities. To assess the performance of model in MESE and facilitate further research, we constructed the MESED dataset which is the first multi-modal dataset for ESE with large-scale and elaborate manual calibration. A powerful multi-modal model MultiExpan is proposed which is pre-trained on four multimodal pre-training tasks. The extensive experiments and analyses on MESED demonstrate the high quality of the dataset and the effectiveness of our MultiExpan, as well as pointing the direction for future research.

Abstract: This paper addresses the problem of selecting of a set of texts for annotation in text classification using retrieval methods when there are limits on the number of annotations due to constraints on human resources. An additional challenge addressed is dealing with binary categories that have a small number of positive instances, reflecting severe class imbalance. In our situation, where annotation occurs over a long time period, the selection of texts to be annotated can be made in batches, with previous annotations guiding the choice of the next set. To address these challenges, the paper proposes leveraging SHAP to construct a quality set of queries for Elasticsearch and semantic search, to try to identify optimal sets of texts for annotation that will help with class imbalance. The approach is tested on sets of cue texts describing possible future events, constructed by participants involved in studies aimed to help with the management of obesity and diabetes. We introduce an effective method for selecting a small set of texts for annotation and building high-quality classifiers. We integrate vector search, semantic search, and machine learning classifiers to yield a good solution. Our experiments demonstrate improved F1 scores for the minority classes in binary classification.

Abstract: Fanfiction, a popular form of creative writing set within established fictional universes, has gained a substantial online following. However, ensuring the well-being and safety of participants has become a critical concern in this community. The detection of triggering content, material that may cause emotional distress or trauma to readers, poses a significant challenge. In this paper, we describe our approach for the Trigger Detection shared task at PAN CLEF 2023, where we want to detect multiple triggering content in a given Fanfiction document. For this, we build a hierarchical model that uses recurrence over Transformer-based language models. In our approach, we first split long documents into smaller sized segments and use them to fine-tune a Transformer model. Then, we extract feature embeddings from the fine-tuned Transformer model, which are used as input in the training of multiple LSTM models for trigger detection in a multi-label setting. Our model achieves an F1-macro score of 0.372 and F1-micro score of 0.736 on the validation set, which are higher than the baseline results shared at PAN CLEF 2023.

Abstract: The task of multi-author writing style detection aims at finding any positions of writing style change in a given text document. We formulate the task as a natural language inference problem where two consecutive paragraphs are paired. Our approach focuses on transitions between paragraphs while truncating input tokens for the task. As backbone models, we employ different Transformer-based encoders with warmup phase during training. We submit the model version that outperforms baselines and other proposed model versions in our experiments. For the easy and medium setups, we submit transition-focused natural language inference based on DeBERTa with warmup training, and the same model without transition for the hard setup.

Abstract: Large Language Models for Code (Code LLM) are flourishing. New and powerful models are released on a weekly basis, demonstrating remarkable performance on the code generation task. Various approaches have been proposed to boost the code generation performance of pre-trained Code LLMs, such as supervised fine-tuning, instruction tuning, reinforcement learning, etc. In this paper, we propose a novel RRTF (Rank Responses to align Test&Teacher Feedback) framework, which can effectively and efficiently boost pre-trained large language models for code generation. Under this framework, we present PanGu-Coder2, which achieves 62.20% pass@1 on the OpenAI HumanEval benchmark. Furthermore, through an extensive evaluation on CoderEval and LeetCode benchmarks, we show that PanGu-Coder2 consistently outperforms all previous Code LLMs.

Abstract: We present TransNormerLLM, the first linear attention-based Large Language Model (LLM) that outperforms conventional softmax attention-based models in terms of both accuracy and efficiency. TransNormerLLM evolves from the previous linear attention architecture TransNormer by making advanced modifications that include positional embedding, linear attention acceleration, gating mechanism, tensor normalization, inference acceleration and stabilization. Specifically, we use LRPE together with an exponential decay to avoid attention dilution issues while allowing the model to retain global interactions between tokens. Additionally, we propose Lightning Attention, a cutting-edge technique that accelerates linear attention by more than twice in runtime and reduces memory usage by a remarkable four times. To further enhance the performance of TransNormer, we leverage a gating mechanism to smooth training and a new tensor normalization scheme to accelerate the model, resulting in an impressive acceleration of over 20%. Furthermore, we have developed a robust inference algorithm that ensures numerical stability and consistent inference speed, regardless of the sequence length, showcasing superior efficiency during both training and inference stages. Scalability is at the heart of our model's design, enabling seamless deployment on large-scale clusters and facilitating expansion to even more extensive models, all while maintaining outstanding performance metrics. Rigorous validation of our model design is achieved through a series of comprehensive experiments on our self-collected corpus, boasting a size exceeding 6TB and containing over 2 trillion tokens. To ensure data quality and relevance, we implement a new self-cleaning strategy to filter our collected data. Our pre-trained models will be released to foster community advancements in efficient LLMs.

Abstract: The effectiveness of compression distance in KNN-based text classification ('gzip') has recently garnered lots of attention. In this note, we show that similar or better effectiveness can be achieved with simpler means, and text compression may not be necessary. Indeed, we find that a simple 'bag-of-words' matching can achieve similar or better accuracy, and is more efficient.

Abstract: Large language models (LLMs) have shown the potential to be integrated into human daily lives. Therefore, user preference is the most critical criterion for assessing LLMs' performance in real-world scenarios. However, existing benchmarks mainly focus on measuring models' accuracy using multi-choice questions, which limits the understanding of their capabilities in real applications. We fill this gap by proposing a comprehensive Chinese benchmark SuperCLUE, named after another popular Chinese LLM benchmark CLUE. SuperCLUE encompasses three sub-tasks: actual users' queries and ratings derived from an LLM battle platform (CArena), open-ended questions with single and multiple-turn dialogues (OPEN), and closed-ended questions with the same stems as open-ended single-turn ones (CLOSE). Our study shows that accuracy on closed-ended questions is insufficient to reflect human preferences achieved on open-ended ones. At the same time, they can complement each other to predict actual user preferences. We also demonstrate that GPT-4 is a reliable judge to automatically evaluate human preferences on open-ended questions in a Chinese context. Our benchmark will be released at https://www.CLUEbenchmarks.com

Abstract: Because "out-of-the-box" large language models are capable of generating a great deal of objectionable content, recent work has focused on aligning these models in an attempt to prevent undesirable generation. While there has been some success at circumventing these measures -- so-called "jailbreaks" against LLMs -- these attacks have required significant human ingenuity and are brittle in practice. In this paper, we propose a simple and effective attack method that causes aligned language models to generate objectionable behaviors. Specifically, our approach finds a suffix that, when attached to a wide range of queries for an LLM to produce objectionable content, aims to maximize the probability that the model produces an affirmative response (rather than refusing to answer). However, instead of relying on manual engineering, our approach automatically produces these adversarial suffixes by a combination of greedy and gradient-based search techniques, and also improves over past automatic prompt generation methods. Surprisingly, we find that the adversarial prompts generated by our approach are quite transferable, including to black-box, publicly released LLMs. Specifically, we train an adversarial attack suffix on multiple prompts (i.e., queries asking for many different types of objectionable content), as well as multiple models (in our case, Vicuna-7B and 13B). When doing so, the resulting attack suffix is able to induce objectionable content in the public interfaces to ChatGPT, Bard, and Claude, as well as open source LLMs such as LLaMA-2-Chat, Pythia, Falcon, and others. In total, this work significantly advances the state-of-the-art in adversarial attacks against aligned language models, raising important questions about how such systems can be prevented from producing objectionable information. Code is available at github.com/llm-attacks/llm-attacks.

Abstract: Clinical trials are vital in advancing drug development and evidence-based medicine, but their success is often hindered by challenges in patient recruitment. In this work, we investigate the potential of large language models (LLMs) to assist individual patients and referral physicians in identifying suitable clinical trials from an extensive selection. Specifically, we introduce TrialGPT, a novel architecture employing LLMs to predict criterion-level eligibility with detailed explanations, which are then aggregated for ranking and excluding candidate clinical trials based on free-text patient notes. We evaluate TrialGPT on three publicly available cohorts of 184 patients and 18,238 annotated clinical trials. The experimental results demonstrate several key findings: First, TrialGPT achieves high criterion-level prediction accuracy with faithful explanations. Second, the aggregated trial-level TrialGPT scores are highly correlated with expert eligibility annotations. Third, these scores prove effective in ranking clinical trials and exclude ineligible candidates. Our error analysis suggests that current LLMs still make some mistakes due to limited medical knowledge and domain-specific context understanding. Nonetheless, we believe the explanatory capabilities of LLMs are highly valuable. Future research is warranted on how such AI assistants can be integrated into the routine trial matching workflow in real-world settings to improve its efficiency.

Abstract: Large language models rely on real-valued representations of text to make their predictions. These representations contain information learned from the data that the model has trained on, including knowledge of linguistic properties and forms of demographic bias, e.g., based on gender. A growing body of work has considered information about concepts such as these using orthogonal projections onto subspaces of the representation space. We contribute to this body of work by proposing a formal definition of intrinsic information in a subspace of a language model's representation space. We propose a counterfactual approach that avoids the failure mode of spurious correlations (Kumar et al., 2022) by treating components in the subspace and its orthogonal complement independently. We show that our counterfactual notion of information in a subspace is optimizing by an causal concept subspace. Furthermore, this intervention allows us to attempt concept controlled generation by manipulating the value of the conceptual component of a representation. Empirically, we find that R-LACE (Ravfogel et al., 2022) returns a one-dimensional subspace containing roughly half of total concept information under our framework. Our causal controlled intervention shows that, for at least one model, the subspace returned by R-LACE can be used to manipulate the concept value of the generated word with precision.