NAACL 2024 Papers

Extensive efforts in the past have been directed toward the development of summarization datasets. However, a predominant number of these resources have been (semi)-automatically generated, typically through web data crawling. This resulted in subpar resources for training and evaluating summarization systems, a quality compromise that is arguably due to the substantial costs associated with generating ground-truth summaries, particularly for diverse languages and specialized domains. To address this issue, we present ACLSum, a novel summarization dataset carefully crafted and evaluated by domain experts. In contrast to previous datasets, ACLSum facilitates multi-aspect summarization of scientific papers, covering challenges, approaches, and outcomes in depth. Through extensive experiments, we evaluate the quality of our resource and the performance of models based on pretrained language models (PLMs) and state-of-the-art large language models (LLMs). Additionally, we explore the effectiveness of extract-then-abstract versus abstractive end-to-end summarization within the scholarly domain on the basis of automatically discovered aspects. While the former performs comparably well to the end-to-end approach with pretrained language models regardless of the potential error propagation issue, the prompting-based approach with LLMs shows a limitation in extracting sentences from source documents.

Babel Briefings is a novel dataset featuring 4.7 million news headlines from August 2020 to November 2021, across 30 languages and 54 locations worldwide with English translations of all articles included. Designed for natural language processing and media studies, it serves as a high-quality dataset for training or evaluating language models as well as offering a simple, accessible collection of articles, for example, to analyze global news coverage and cultural narratives. As a simple demonstration of the analyses facilitated by this dataset, we use a basic procedure using a TF-IDF weighted similarity metric to group articles into clusters about the same event. We then visualize the event signatures of the event showing articles of which languages appear over time, revealing intuitive features based on the proximity of the event and unexpectedness of the event. The dataset is available on [Kaggle](https://www.kaggle.com/datasets/felixludos/babel-briefings) and [HuggingFace](https://huggingface.co/datasets/felixludos/babel-briefings) with accompanying [GitHub](https://github.com/felixludos/babel-briefings) code.

Systematic Reviews (SRs) are foundational in healthcare for synthesising evidence to inform clinical practices. Traditionally skewed towards English-language databases, SRs often exclude significant research in other languages, leading to potential biases. This study addresses this gap by focusing on Spanish, a language notably underrepresented in SRs. We present a foundational zero-shot dual information retrieval (IR) baseline system, integrating traditional retrieval methods with pre-trained language models and cross-attention re-rankers for enhanced accuracy in Spanish biomedical literature retrieval. Utilising the LILACS database, known for its comprehensive coverage of Latin American and Caribbean biomedical literature, we evaluate the approach with three real-life case studies in Spanish SRs. The findings demonstrate the system’s efficacy and underscore the importance of query formulation. This study contributes to the field of IR by promoting language inclusivity and supports the development of more comprehensive and globally representative healthcare guidelines.

Large Language Models (LLMs), such as ChatGPT and GPT-4, are designed to provide useful and safe responses. However, adversarial prompts known as ‘jailbreaks’ can circumvent safeguards, leading LLMs to generate potentially harmful content. Exploring jailbreak prompts can help to better reveal the weaknesses of LLMs and further steer us to secure them. Unfortunately, existing jailbreak methods either suffer from intricate manual design or require optimization on other white-box models, which compromises either generalization or efficiency. In this paper, we generalize jailbreak prompt attacks into two aspects: (1) Prompt Rewriting and (2) Scenario Nesting. Based on this, we propose ReNeLLM, an automatic framework that leverages LLMs themselves to generate effective jailbreak prompts. Extensive experiments demonstrate that ReNeLLM significantly improves the attack success rate while greatly reducing the time cost compared to existing baselines. Our study also reveals the inadequacy of current defense methods in safeguarding LLMs. Finally, we analyze the failure of LLMs defense from the perspective of prompt execution priority, and propose corresponding defense strategies. We hope that our research can catalyze both the academic community and LLMs developers towards the provision of safer and more regulated LLMs. The code is available at https://github.com/NJUNLP/ReNeLLM.

We present a Universal Dependencies (UD) treebank for Highland Puebla Nahuatl. The treebank is only the second such UD corpus for a Mexican language, and supplements an existing treebank for another Nahuatl variant. We describe the process of data collection, annotation decisions and interesting syntactic constructions, and discuss some similarities and differences between the Highland Puebla Nahuatl treebank and the existing Western Sierra Puebla Nahuatl treebank.

Reasoning methods, best exemplified by the well-known Chain-of-Thought (CoT), empower the reasoning abilities of Large Language Models (LLMs) by eliciting them to solve complex tasks in a step-by-step manner. Although they are achieving significant success, the ability to deliver multi-step reasoning remains limited to English because of the imbalance in the distribution of pre-training data, which makes other languages a barrier. In this paper, we propose Cross-lingual Tree-of-Thoughts (Cross-ToT), a method for aligning Cross-lingual CoT reasoning across languages. The proposed method, through a self-consistent cross-lingual prompting mechanism inspired by the Tree-of-Thoughts approach, provides multi-step reasoning paths in different languages that, during the steps, lead to the final solution. Experimental evaluations show that our method significantly outperforms existing prompting methods by reducing the number of interactions and achieving state-of-the-art performance.

Natural languages are believed to be (mildly) context-sensitive. Despite underpinning remarkably capable large language models, transformers are unable to model many context-free language tasks. In an attempt to address this limitation in the modeling power of transformer-based language models, we propose augmenting them with a differentiable, stack-based attention mechanism. Our stack-basedattention mechanism can be incorporated into any transformer-based language model and adds a level of interpretability to the model. We show that the addition of our stack-based attention mechanism enables the transformer to model some, but not all, deterministic context-freelanguages.

Metaphor detection is a challenging task in figurative language processing, which aims to distinguish between metaphorical and literal expressions in text. Existing methods tackle metaphor detection via training or fine-tuning discriminative models on labeled data. However, these approaches struggle to explain the underlying reasoning process behind the metaphorical/literal judgment. Recently, large language models (LLMs) have shown promise in language reasoning tasks. Although promising, LLM-based methods for metaphor detection and reasoning are still faced with the challenging issue of bringing the explainable concepts for metaphor reasoning and their linguistic manifestation. To fill this gap, we propose a novel Theory guided Scaffolding Instruction (TSI) framework that instructs an LLM to infer the underlying reasoning process of metaphor detection guided by metaphor theories for the first time. Our work is inspired by a pedagogical strategy called scaffolding instruction, which encourages educators to provide questioning and support as scaffolding so as to assist learners in constructing the understanding of pedagogical goals step by step. We first construct a metaphor knowledge graph grounded in metaphor theory which serves as the instructional structure to obtain a series of scaffolding questions, directing the LLM to incrementally generate the reasoning process for metaphor understanding through dialogue interactions. During this theory guided instruction process, we explore the LLM’s mastery boundary and provide the relevant knowledge as scaffolding support when the question is beyond the LLM’s capability. Experimental results verify that our method significantly outperforms both the LLM-based reasoning methods and the SOTA methods in metaphor detection, indicating the facilitation of metaphor and instruction theories in guiding LLM-based reasoning process.

A central component of rational behavior is logical inference: the process of determining which conclusions follow from a set of premises. Psychologists have documented several ways in which humans’ inferences deviate from the rules of logic. Do language models, which are trained on text generated by humans, replicate such human biases, or are they able to overcome them? Focusing on the case of syllogisms—inferences from two simple premises—we show that, within the PaLM 2 family of transformer language models, larger models are more logical than smaller ones, and also more logical than humans. At the same time, even the largest models make systematic errors, some of which mirror human reasoning biases: they show sensitivity to the (irrelevant) ordering of the variables in the syllogism, and draw confident but incorrect inferences from particular syllogisms (syllogistic fallacies). Overall, we find that language models often mimic the human biases included in their training data, but are able to overcome them in some cases.

Contextualized embeddings are the preferred tool for modeling Lexical Semantic Change (LSC). Current evaluations typically focus on a specific task known as Graded Change Detection (GCD). However, performance comparison across work are often misleading due to their reliance on diverse settings. In this paper, we evaluate state-of-the-art models and approaches for GCD under equal conditions. We further break the LSC problem into Word-in-Context (WiC) and Word Sense Induction (WSI) tasks, and compare models across these different levels. Our evaluation is performed across different languages on eight available benchmarks for LSC, and shows that (i) APD outperforms other approaches for GCD; (ii) XL-LEXEME outperforms other contextualized models for WiC, WSI, and GCD, while being comparable to GPT-4; (iii) there is a clear need for improving the modeling of word meanings, as well as focus on *how*, *when*, and *why* these meanings change, rather than solely focusing on the extent of semantic change.

Multilingual modelling can improve machine translation for low-resource languages, partly through shared subword representations. This paper studies the role of subword segmentation in cross-lingual transfer. We systematically compare the efficacy of several subword methods in promoting synergy and preventing interference across different linguistic typologies. Our findings show that subword regularisation boosts synergy in multilingual modelling, whereas BPE more effectively facilitates transfer during cross-lingual fine-tuning. Notably, our results suggest that differences in orthographic word boundary conventions (the morphological granularity of written words) may impede cross-lingual transfer more significantly than linguistic unrelatedness. Our study confirms that decisions around subword modelling can be key to optimising the benefits of multilingual modelling.

This paper proposes a methodology for generating and perturbing detailed derivations of equations at scale, aided by a symbolic engine, to evaluate the generalisability of Transformers to out-of-distribution mathematical reasoning problems. Instantiating the framework in the context of sequence classification tasks, we compare the capabilities of GPT-4, GPT-3.5, and a canon of fine-tuned BERT models, exploring the relationship between specific operators and generalisation failure via the perturbation of reasoning aspects such as symmetry and variable surface forms. Surprisingly, our empirical evaluation reveals that the average in-distribution performance of fine-tuned models surpasses GPT-3.5, and rivals GPT-4. However, perturbations to input reasoning can reduce their performance by up to 80 F1 points. Overall, the results suggest that the in-distribution performance of smaller open-source models may potentially rival GPT by incorporating appropriately structured derivation dependencies during training, and highlight a shared weakness between BERT and GPT involving a relative inability to decode indirect references to mathematical entities. We release the full codebase, constructed datasets, and fine-tuned models to encourage future progress in the field.

Symbolic meaning representations of natural language text have been studied since at least the 1960s. With the availability of large annotated corpora, and more powerful machine learning tools, the field has recently seen several new developments. In this survey, we study today’s most prominent Meaning Representation Frameworks. We shed light on their theoretical properties, as well as on their practical research environment, i.e., on datasets, parsers, applications, and future challenges.

Large language models (LLMs) have demonstrated remarkable capabilities across a wide range of tasks in various domains. Despite their impressive performance, they can be unreliable due to factual errors in their generations. Assessing their confidence and calibrating them across different tasks can help mitigate risks and enable LLMs to produce better generations. There has been a lot of recent research aiming to address this, but there has been no comprehensive overview to organize it and to outline the main lessons learned. The present survey aims to bridge this gap. In particular, we outline the challenges and we summarize recent technical advancements for LLM confidence estimation and calibration. We further discuss their applications and suggest promising directions for future work.

Accurate uncertainty quantification is crucial for the safe deployment of machine learning models, and prior research has demonstrated improvements in the calibration of modern language models (LMs). We study in-context learning (ICL), a prevalent method for adapting static LMs through tailored prompts, and examine the balance between performance and calibration across a broad spectrum of natural language understanding and reasoning tasks. Through comprehensive experiments, we observe that, with an increasing number of ICL examples, models initially exhibit increased miscalibration before achieving better calibration and miscalibration tends to arise in low-shot settings. Moreover, we find that methods aimed at improving usability, such as fine-tuning and chain-of-thought (CoT) prompting, can lead to miscalibration and unreliable natural language explanations. Furthermore, we explore recalibration techniques and find that a scaling-binning calibrator can reduce calibration errors consistently.

Media framing is the study of strategically selecting and presenting specific aspects of political issues to shape public opinion. Despite its relevance to almost all societies around the world, research has been limited due to the lack of available datasets and other resources. This study explores the possibility of dataset creation through crowdsourcing, utilizing non-expert annotators to develop training corpora. We first extend framing analysis beyond English news to a multilingual context (12 typologically diverse languages) through automatic translation. We also present a novel benchmark in Bengali and Portuguese on the immigration and same-sex marriage domains.Additionally, we show that a system trained on our crowd-sourced dataset, combined with other existing ones, leads to a 5.32 percentage point increase from the baseline, showing that crowdsourcing is a viable option. Last, we study the performance of large language models (LLMs) for this task, finding that task-specific fine-tuning is a better approach than employing bigger non-specialized models.

Learning argumentative writing is challenging. Besides writing fundamentals such as syntax and grammar, learners must select and arrange argument components meaningfully to create high-quality essays. To support argumentative writing computationally, one step is to mine the argumentative structure. When combined with automatic essay scoring, interactions of the argumentative structure and quality scores can be exploited for comprehensive writing support. Although studies have shown the usefulness of using information about the argumentative structure for essay scoring, no argument mining corpus with ground-truth essay quality annotations has been published yet. Moreover, none of the existing corpora contain essays written by school students specifically. To fill this research gap, we present a German corpus of 1,320 essays from school students of two age groups. Each essay has been manually annotated for argumentative structure and quality on multiple levels of granularity. We propose baseline approaches to argument mining and essay scoring, and we analyze interactions between both tasks, thereby laying the ground for quality-oriented argumentative writing support.

Large language models (LLMs) have show their remarkable ability in various natural language tasks. However, there are concerns that LLMs are possible to be used improperly or even illegally. To prevent the malicious usage of LLMs, detecting LLM-generated text becomes crucial in the deployment of LLM applications. Watermarking is an effective strategy to detect the LLM-generated content by encoding a pre-defined secret watermark to facilitate the detection process. However, the majority of existing watermark methods leverage the simple hashes of precedent tokens to partition vocabulary. Such watermarks can be easily eliminated by paraphrase and, correspondingly, the detection effectiveness will be greatly compromised. Thus, to enhance the robustness against paraphrase, we propose a semantics-based watermark framework, SemaMark. It leverages the semantics as an alternative to simple hashes of tokens since the semantic meaning of the sentences will be likely preserved under paraphrase and the watermark can remain robust. Comprehensive experiments are conducted to demonstrate the effectiveness and robustness of SemaMark under different paraphrases.

Based on Pre-trained Language Models (PLMs), event coreference resolution (ECR) systems have demonstrated outstanding performance in clustering coreferential events across documents. However, the state-of-the-art system exhibits an excessive reliance on the ‘triggers lexical matching’ spurious pattern in the input mention pair text. We formalize the decision-making process of the baseline ECR system using a Structural Causal Model (SCM), aiming to identify spurious and causal associations (i.e., rationales) within the ECR task. Leveraging the debiasing capability of counterfactual data augmentation, we develop a rationale-centric counterfactual data augmentation method with LLM-in-the-loop. This method is specialized for pairwise input in the ECR system, where we conduct direct interventions on triggers and context to mitigate the spurious association while emphasizing the causation. Our approach achieves state-of-the-art performance on three popular cross-document ECR benchmarks and demonstrates robustness in out-of-domain scenarios.