EMNLP 2024 论文列表

Legal case retrieval for sourcing similar cases is critical in upholding judicial fairness. Different from general web search, legal case retrieval involves processing lengthy, complex, and highly specialized legal documents. Existing methods in this domain often overlook the incorporation of legal expert knowledge, which is crucial for accurately understanding and modeling legal cases, leading to unsatisfactory retrieval performance. This paper introduces KELLER, a legal knowledge-guided case reformulation approach based on large language models (LLMs) for effective and interpretable legal case retrieval. By incorporating professional legal knowledge about crimes and law articles, we enable large language models to accurately reformulate the original legal case into concise sub-facts of crimes, which contain the essential information of the case. Extensive experiments on two legal case retrieval benchmarks demonstrate superior retrieval performance and robustness on complex legal case queries of KELLER over existing methods.

Personalized product search aims to learn personalized preferences from search logs and adjust the ranking lists returned by engines. Previous studies have extensively explored excavating valuable features to build accurate interest profiles. However, they overlook that the user’s attention varies on product attributes(e.g., brand, category). Users may especially prefer specific attributes or switch their preferences between attributes dynamically. Instead, existing approaches mix up all attribute features and let the model automatically extract useful ones from rather complex scenarios. To solve this problem, in this paper, we propose a dynamic multi-attribute interest learning model to tackle the influences from attributes to user interests. Specifically, we design two interest profiling modules: attribute-centered and attribute-aware profiling. The former focuses on capturing the user’s preferences on a single attribute, while the latter focuses on addressing the interests correlated with multi-attribute within the search history. Besides, we devise a dynamic contribution weights strategy that sends explicit signals to the model to determine the impacts of different attributes better. Experimental results on large-scale datasets illustrate that our model significantly improves the results of existing methods.

Traditional autonomous driving systems have mainly focused on making driving decisions without human interaction, overlooking human-like decision-making and human preference required in complex traffic scenarios. To bridge this gap, we introduce a novel framework leveraging Large Language Models (LLMs) for learning human-centered driving decisions from diverse simulation scenarios and environments that incorporate human feedback. Our contributions include a GPT-4-based programming planner that integrates seamlessly with the existing CARLA simulator to understand traffic scenes and react to human instructions. Specifically, we build a human-guided learning pipeline that incorporates human driver feedback directly into the learning process and stores optimal driving programming policy using Retrieval Augmented Generation (RAG). Impressively, our programming planner, with only 50 saved code snippets, can match the performance of baseline extensively trained reinforcement learning (RL) models. Our paper highlights the potential of an LLM-powered shared-autonomy system, pushing the frontier of autonomous driving system development to be more interactive and intuitive.

Large language models (LLMs) have demonstrated impressive language understanding and generation capabilities, enabling them to answer a wide range of questions across various domains. However, these models are not flawless and often produce responses that contain errors or misinformation. These inaccuracies, commonly referred to as hallucinations, render LLMs unreliable and even unusable in many scenarios. In this paper, our focus is on mitigating the issue of hallucination in LLMs, particularly in the context of question-answering. Instead of attempting to answer all questions, we explore a refusal mechanism that instructs LLMs to refuse to answer challenging questions in order to avoid errors. We then propose a simple yet effective solution called Learn to Refuse (L2R), which incorporates the refusal mechanism to enable LLMs to recognize and refuse to answer questions that they find difficult to address. To achieve this, we utilize a structured knowledge base to represent all the LLM’s understanding of the world, enabling it to provide traceable gold knowledge. This knowledge base is separate from the LLM and initially empty. It can be filled with validated knowledge and progressively expanded. When an LLM encounters questions outside its domain, the system recognizes its knowledge scope and determines whether it can answer the question independently. Additionally, we introduce a method for automatically and efficiently expanding the knowledge base of LLMs. Through qualitative and quantitative analysis, we demonstrate that our approach enhances the controllability and reliability of LLMs.

Supervised fine-tuning enhances the problem-solving abilities of language models across various mathematical reasoning tasks. To maximize such benefits, existing research focuses on *broadening* the training set with various data augmentation techniques, which is effective for standard single-round question-answering settings. Our work introduces a novel technique aimed at cultivating a *deeper* understanding of the training problems at hand, enhancing performance not only in standard settings but also in more complex scenarios that require reflective thinking. Specifically, we propose **reflective augmentation**, a method that embeds problem reflection into each training instance. It trains the model to consider alternative perspectives and engage with abstractions and analogies, thereby fostering a thorough comprehension through reflective reasoning. Extensive experiments validate the achievement of our aim, underscoring the unique advantages of our method and its complementary nature relative to existing augmentation techniques.

Parameter-Efficient Fine-Tuning (PEFT) methods have gained significant popularity for adapting pre-trained Large Language Models (LLMs) to downstream tasks, primarily due to their potential to significantly reduce memory and computational overheads. However, a common limitation in most PEFT approaches is their application of a uniform architectural design across all layers. This uniformity involves identical trainable modules and ignores the varying importance of each layer, leading to sub-optimal fine-tuning results. To overcome the above limitation and obtain better performance, we develop a novel approach, Importance-aware Sparse Tuning (IST), to fully utilize the inherent sparsity and select the most important subset of full layers with effective layer-wise importance scoring. The proposed IST is a versatile and plug-and-play technique compatible with various PEFT methods that operate on a per-layer basis. By leveraging the estimated importance scores, IST dynamically updates these selected layers in PEFT modules, leading to reduced memory demands. We further provide theoretical proof of convergence and empirical evidence of superior performance to demonstrate the advantages of IST over uniform updating strategies. Extensive experiments on a range of LLMs, PEFTs, and downstream tasks substantiate the effectiveness of our proposed method, showcasing IST’s capacity to enhance existing layer-based PEFT methods. Our code is available at https://github.com/Kaiseem/IST

Fine-tuning pre-trained large language models (LLMs) on a diverse array of tasks has become a common approach for building models that can solve various natural language processing (NLP) tasks. However, where and to what extent these models retain task-specific knowledge remains largely unexplored. This study investigates the task-specific information encoded in pre-trained LLMs and the effects of instruction tuning on their representations across a diverse set of over 60 NLP tasks. We use a set of matrix analysis tools to examine the differences between the way pre-trained and instruction-tuned LLMs store task-specific information. Our findings reveal that while some tasks are already encoded within the pre-trained LLMs, others greatly benefit from instruction tuning. Additionally, we pinpointed the layers in which the model transitions from high-level general representations to more task-oriented representations. This finding extends our understanding of the governing mechanisms of LLMs and facilitates future research in the fields of parameter-efficient transfer learning and multi-task learning. Our code is available at: https://github.com/zsquaredz/layer_by_layer/

We present LawBench, the first evaluation benchmark composed of 20 tasks aimed to assess the ability of Large Language Models (LLMs) to perform Chinese legal-related tasks. LawBench is meticulously crafted to enable precise assessment of LLMs’ legal capabilities from three cognitive levels that correspond to the widely accepted Bloom’s cognitive taxonomy. Using LawBench, we present a comprehensive evaluation of 21 popular LLMs and the first comparative analysis of the empirical results in order to reveal their relative strengths and weaknesses. All data, model predictions and evaluation code are accessible from https://github.com/open-compass/LawBench.

Interpreting and understanding the predictions made by deep learning models poses a formidable challenge due to their inherently opaque nature. Many previous efforts aimed at explaining these predictions rely on input features, specifically, the words within NLP models. However, such explanations are often less informative due to the discrete nature of these words and their lack of contextual verbosity. To address this limitation, we introduce the Latent Concept Attribution method (LACOAT), which generates explanations for predictions based on latent concepts. Our foundational intuition is that a word can exhibit multiple facets, contingent upon the context in which it is used. Therefore, given a word in context, the latent space derived from our training process reflects a specific facet of that word. LACOAT functions by mapping the representations of salient input words into the training latent space, allowing it to provide latent context-based explanations of the prediction.

The use of propagandistic techniques in online content has increased in recent years aiming to manipulate online audiences. Fine-grained propaganda detection and extraction of textual spans where propaganda techniques are used, are essential for more informed content consumption. Automatic systems targeting the task over lower resourced languages are limited, usually obstructed by lack of large scale training datasets. Our study investigates whether Large Language Models (LLMs), such as GPT-4, can effectively extract propagandistic spans. We further study the potential of employing the model to collect more cost-effective annotations. Finally, we examine the effectiveness of labels provided by GPT-4 in training smaller language models for the task. The experiments are performed over a large-scale in-house manually annotated dataset. The results suggest that providing more annotation context to GPT-4 within prompts improves its performance compared to human annotators. Moreover, when serving as an expert annotator (consolidator), the model provides labels that have higher agreement with expert annotators, and lead to specialized models that achieve state-of-the-art over an unseen Arabic testing set. Finally, our work is the first to show the potential of utilizing LLMs to develop annotated datasets for propagandistic spans detection task prompting it with annotations from human annotators with limited expertise. All scripts and annotations will be shared with the community.

Data annotation and synthesis generally refers to the labeling or generating of raw data with relevant information, which could be used for improving the efficacy of machine learning models. The process, however, is labor-intensive and costly. The emergence of advanced Large Language Models (LLMs), exemplified by GPT-4, presents an unprecedented opportunity to automate the complicated process of data annotation and synthesis. While existing surveys have extensively covered LLM architecture, training, and general applications, we uniquely focus on their specific utility for data annotation. This survey contributes to three core aspects: LLM-Based Annotation Generation, LLM-Generated Annotations Assessment, and LLM-Generated Annotations Utilization. Furthermore, this survey includes an in-depth taxonomy of data types that LLMs can annotate, a comprehensive review of learning strategies for models utilizing LLM-generated annotations, and a detailed discussion of the primary challenges and limitations associated with using LLMs for data annotation and synthesis. Serving as a key guide, this survey aims to assist researchers and practitioners in exploring the potential of the latest LLMs for data annotation, thereby fostering future advancements in this critical field.

Pre-trained language models like BERT and T5 serve as crucial backbone encoders for dense retrieval. However, these models often exhibit limited generalization capabilities and face challenges in improving in-domain accuracy. Recent research has explored using large language models (LLMs) as retrievers, achieving state-of-the-art performance across various tasks. Despite these advancements, the specific benefits of LLMs over traditional retrievers and the impact of different LLM configurations—such as parameter sizes, pre-training duration, and alignment processes—on retrieval tasks remain unclear. In this work, we conduct a comprehensive empirical study on a wide range of retrieval tasks, including in-domain accuracy, data efficiency, zero-shot generalization, lengthy retrieval, instruction-based retrieval, and multi-task learning. We evaluate over 15 different backbone LLMs and non-LLMs. Our findings reveal that larger models and extensive pre-training consistently enhance in-domain accuracy and data efficiency. Additionally, larger models demonstrate significant potential in zero-shot generalization, lengthy retrieval, instruction-based retrieval, and multi-task learning. These results underscore the advantages of LLMs as versatile and effective backbone encoders in dense retrieval, providing valuable insights for future research and development in this field.

Recent advancements in educational platforms have emphasized the importance of personalized education. Accurately estimating question difficulty based on the ability of the student group is essential for personalized question recommendations. Several studies have focused on predicting question difficulty using student question-solving records or textual information about the questions. However, these approaches require a large amount of student question-solving records and fail to account for the subjective difficulties perceived by different student groups. To address these limitations, we propose the LLaSA framework that utilizes large language models to represent students at various levels. Our proposed method, LLaSA and the zero-shot LLaSA, can estimate question difficulty both with and without students’ question-solving records. In evaluations on the DBE-KT22 and ASSISTMents 2005–2006 benchmarks, the zero-shot LLaSA demonstrated a performance comparable to those of strong baseline models even without any training. When evaluated using the classification method, LLaSA outperformed the baseline models, achieving state-of-the-art performance. In addition, the zero-shot LLaSA showed a high correlation with the regressed IRT curve when compared to question difficulty derived from students’ question-solving records, highlighting its potential for real-world applications.

Large Language Models (LLMs) possess extensive knowledge and strong capabilities in performing in-context reasoning. However, previous work challenges their out-of-context reasoning ability, i.e., the ability to infer information from their training data, instead of from the context or prompt. This paper focuses on a significant aspect of out-of-context reasoning: Out-of-Context Knowledge Reasoning (OCKR), which is to combine multiple knowledge to infer new knowledge. We designed a synthetic dataset with seven representative OCKR tasks to systematically assess the OCKR capabilities of LLMs. Using this dataset, we evaluated several LLMs and discovered that their proficiency in this aspect is limited, regardless of whether the knowledge is trained in a separate or adjacent training settings. Moreover, training the model to reason with reasoning examples does not result in significant improvement, while training the model to perform explicit knowledge retrieval helps for retrieving attribute knowledge but not the relation knowledge, indicating that the model’s limited OCKR capabilities are due to difficulties in knowledge retrieval. Furthermore, we treat cross-lingual knowledge transfer as a distinct form of OCKR, and evaluate this ability. Our results show that the evaluated model also exhibits limited ability in transferring knowledge across languages.

In this work, we study in-context teaching(ICT), where a teacher provides in-context example rationales to teach a student to reasonover unseen cases. Human teachers are usually required to craft in-context demonstrations, which are costly and have high variance. We ask whether a large language model (LLM) can serve as a more effective in-context teacher for itself or otherLLMs, compared to humans. Inspired by the Encoding Specificity Hypothesis from human episodic memory, we hypothesize thatin-context exemplars crafted by the teacher should match the training data of the student. This hypothesis motivates us to propose Self-Explain where an LLM’s self-elicited explanations are used as in-context demonstrations for prompting it as they are generalized fromthe model’s training examples. Self-Explain is shown to significantly outperform using human-crafted exemplars and other baselines.Furthermore, we reveal that for ICT, rationales from different teacher LLMs or human experts that more resemble the student LLM’s self-explanations are better in-context demonstrations. This supports our encoding specificity hypothesis. We then propose Teach-Back that aligns a teacher LLM with the student to enhance the ICT performance. For example, Teach-Back enables a 7B model to teach the much larger GPT-3.5 in context, surpassing human teachers by around 5% in test accuracy on medical question answering.

Visual question answering (VQA) tasks, often performed by visual language model (VLM), face challenges with long-tail knowledge. Recent retrieval-augmented VQA (RA-VQA) systems address this by retrieving and integrating external knowledge sources. However, these systems still suffer from redundant visual information irrelevant to the question during retrieval. To address these issues, in this paper, we propose LLM-RA, a novel method leveraging the reasoning capability of a large language model (LLM) to identify key visual entities, thus minimizing the impact of irrelevant information in the query of retriever. Furthermore, key visual entities are independently encoded for multimodal joint retrieval, preventing cross-entity interference. Experimental results demonstrate that our method outperforms other strong RA-VQA systems. In two knowledge-intensive VQA benchmarks, our method achieves the new state-of-the-art performance among those with similar scale of parameters and even performs comparably to models with 1-2 orders larger parameters.

Turn-taking is a fundamental mechanism in human communication that ensures smooth and coherent verbal interactions. Recent advances in Large Language Models (LLMs) have motivated their use in improving the turn-taking capabilities of Spoken Dialogue Systems (SDS), such as their ability to respond at appropriate times. However, existing models often struggle to predict opportunities for speaking — called Transition Relevance Places (TRPs) — in natural, unscripted conversations, focusing only on turn-final TRPs and not within-turn TRPs. To address these limitations, we introduce a novel dataset of participant-labeled within-turn TRPs and use it to evaluate the performance of state-of-the-art LLMs in predicting opportunities for speaking. Our experiments reveal the current limitations of LLMs in modeling unscripted spoken interactions, highlighting areas for improvement and paving the way for more naturalistic dialogue systems.

Intrinsic self-correct was a method that instructed large language models (LLMs) to verify and correct their responses without external feedback. Unfortunately, the study concluded that the LLMs could not self-correct reasoning yet. We find that a simple yet effective prompting method enhances LLM performance in identifying and correcting inaccurate answers without external feedback.That is to mask a key condition in the question, add the current response to construct a verification question, and predict the condition to verify the response. The condition can be an entity in an open-domain question or a numerical value in an arithmetic question, which requires minimal effort (via prompting) to identify. We propose an iterative verify-then-correct framework to progressively identify and correct (probably) false responses, named ProCo. We conduct experiments on three reasoning tasks. On average, ProCo, with GPT-3.5-Turbo-1106 as the backend LLM, yields +6.8 exact match on four open-domain question answering datasets, +14.1 accuracy on three arithmetic reasoning datasets, and +9.6 accuracy on a commonsense reasoning dataset, compared to Self-Correct.Our implementation is made publicly available at https://wzy6642.github.io/proco.github.io/.

Despite their promising performance across various tasks, recent studies reveal that Large language models (LLMs) still exhibit significant deficiencies in handling several word-level and character-level tasks, e.g., word unscrambling and sentence editing, indicating urgent needs for substantial improvements in basic language understanding and manipulation. To address these challenges, it is crucial to develop large-scale benchmarks that can comprehensively assess the performance of LLMs in basic language tasks. In this paper, we introduce a bilingual benchmark, CWUM, to investigate the capabilities and limitations of LLMs in understanding and manipulating natural language at both character and word levels. CWUM consists of 15 simple text editing tasks, e.g., letter counting, word reversing, Chinese character inserting, etc. We conduct extensive experiments on eight advanced LLMs, including base models and instruction-tuned (chat) variants. The experimental results highlight significant failures of existing LLMs on CWUM tasks that humans can solve perfectly with 100% accuracy. On English tasks of CWUM, the average accuracy of GPT-4, LLaMA-3-70B, and Qwen-72B is 66.64%, 39.32%, and 33.16%, respectively, which lags far behind human performance. Instruction-tuning the base model does not lead to a distinct performance improvement, as the average accuracy of LLaMA-3-70B-Instruct on English tasks is only 1.44% higher than that of the base LLaMA-3-70B. Ultimately, we show that supervised fine-tuning (SFT) can enhance model performance on CWUM without compromising its ability to generalize across general tasks.