Huiyuan Xie

Yuxiao Ye, Yiwen Zhang, Huiyuan Xie et al.

LLM-based multi-agent systems are increasingly used for strategic decision-making tasks. In such settings, performance depends not only on individual model capabilities, but also on the policies by which agents interact and adapt. Multi-agent reinforcement learning can optimise these interaction policies, but its reward design often remains task-specific and weakly grounded in interaction structure. To address this gap, we propose GARL, a GAme-theoretic Reinforcement Learning framework for multi-agent strategic prioritisation. GARL formalises strategic prioritisation as a two-stage game: competing agents first allocate strategic resources over a shared candidate set, and a higher-level arbiter then produces the final ranking. The resulting game-theoretic utilities are converted into role-specific reinforcement signals, allowing policy optimisation to be guided by structured interaction. We instantiate GARL on issues-in-dispute ranking, where the goal is to prioritise core issues in legal proceedings. Experiments show that GARL improves ranking performance, enables small open-source LLMs to become competitive with a strong closed-source LLM under the same candidate-ranking setting, and yields gains in legal-domain competence and broader strategic decision-making. Overall, GARL demonstrates how game-theoretic interaction structure can be turned into reinforcement-learning objectives, providing a principled approach to policy optimisation in multi-agent strategic prioritisation.

Andreas Östling, Holli Sargeant, Huiyuan Xie et al.

We introduce the Cambridge Law Corpus (CLC), a dataset for legal AI research. It consists of over 250 000 court cases from the UK. Most cases are from the 21st century, but the corpus includes cases as old as the 16th century. This paper presents the first release of the corpus, containing the raw text and meta-data. Together with the corpus, we provide annotations on case outcomes for 638 cases, done by legal experts. Using our annotated data, we have trained and evaluated case outcome extraction with GPT-3, GPT-4 and RoBERTa models to provide benchmarks. We include an extensive legal and ethical discussion to address the potentially sensitive nature of this material. As a consequence, the corpus will only be released for research purposes under certain restrictions.

Huiyuan Xie, Felix Steffek, Joana Ribeiro de Faria et al.

This paper explores the intersection of technological innovation and access to justice by developing a benchmark for predicting case outcomes in the UK Employment Tribunal (UKET). To address the challenge of extensive manual annotation, the study employs a large language model (LLM) for automatic annotation, resulting in the creation of the CLC-UKET dataset. The dataset consists of approximately 19,000 UKET cases and their metadata. Comprehensive legal annotations cover facts, claims, precedent references, statutory references, case outcomes, reasons and jurisdiction codes. Facilitated by the CLC-UKET data, we examine a multi-class case outcome prediction task in the UKET. Human predictions are collected to establish a performance reference for model comparison. Empirical results from baseline models indicate that finetuned transformer models outperform zero-shot and few-shot LLMs on the UKET prediction task. The performance of zero-shot LLMs can be enhanced by integrating task-related information into few-shot examples. We hope that the CLC-UKET dataset, along with human annotations and empirical findings, can serve as a valuable benchmark for employment-related dispute resolution.

Yiran Rex Ma, Yuxiao Ye, Huiyuan Xie

Legal text generated by large language models (LLMs) can usually achieve reasonable factual accuracy, but it frequently fails to adhere to the specialised stylistic norms and linguistic conventions of legal writing. In order to improve stylistic quality, a crucial first step is to establish a reliable evaluation method. However, having legal experts manually develop such a metric is impractical, as the implicit stylistic requirements in legal writing practice are difficult to formalise into explicit rubrics. Meanwhile, existing automatic evaluation methods also fall short: reference-based metrics conflate semantic accuracy with stylistic fidelity, and LLM-as-a-judge evaluations suffer from opacity and inconsistency. To address these challenges, we introduce CLASE (Chinese LegAlese Stylistic Evaluation), a hybrid evaluation method that focuses on the stylistic performance of legal text. The method incorporates a hybrid scoring mechanism that combines 1) linguistic feature-based scores and 2) experience-guided LLM-as-a-judge scores. Both the feature coefficients and the LLM scoring experiences are learned from contrastive pairs of authentic legal documents and their LLM-restored counterparts. This hybrid design captures both surface-level features and implicit stylistic norms in a transparent, reference-free manner. Experiments on 200 Chinese legal documents show that CLASE achieves substantially higher alignment with human judgments than traditional metrics and pure LLM-as-a-judge methods. Beyond improved alignment, CLASE provides interpretable score breakdowns and suggestions for improvements, offering a scalable and practical solution for professional stylistic evaluation in legal text generation (Code and data for CLASE is available at: https://github.com/rexera/CLASE).

Yuxiao Ye, Yiming Zhang, Yiran Ma et al.

Large Language Models (LLMs) have enabled Multi-Agent Systems (MASs) where agents interact through natural language to solve complex tasks or simulate multi-party dialogues. Recent work on LLM-based MASs has mainly focused on architecture design, such as role assignment and workflow orchestration. In contrast, this paper targets the interaction process itself, aiming to improve agents' communication efficiency by helping them convey their intended meaning more effectively through language. To this end, we propose LinguaGame, a linguistically-grounded game-theoretic paradigm for multi-agent dialogue generation. Our approach models dialogue as a signalling game over communicative intents and strategies, solved with a training-free equilibrium approximation algorithm for inference-time decision adjustment. Unlike prior game-theoretic MASs, whose game designs are often tightly coupled with task-specific objectives, our framework relies on linguistically informed reasoning with minimal task-specific coupling. Specifically, it treats dialogue as intentional and strategic communication, requiring agents to infer what others aim to achieve (intents) and how they pursue those goals (strategies). We evaluate our framework in simulated courtroom proceedings and debates, with human expert assessments showing significant gains in communication efficiency.

Shuliang Liu, Zhipeng Xu, Zhenghao Liu et al.

Large Language Models (LLMs) as automatic evaluators, commonly referred to as LLM-as-a-Judge, have also attracted growing attention. This approach plays a vital role in aligning LLMs with human judgments, providing accurate and reliable assessments. However, LLM-based judgment models often exhibit judgment preference bias during the evaluation phase, tending to favor responses generated by themselves, undermining the reliability of their judgments. This paper introduces the Group-Based Polling Optimization (Genii), an unsupervised multi-agent collaborative optimization framework that mitigates the inherent judgment preference bias of judgment models. Specifically, Genii integrates various LLM-based judgment models into a multi-agent system and simulates the interactive client-server polling mechanism to optimize each client agent unsupervisedly. Our experiments demonstrate that Genii outperforms supervised models trained on annotated judgment data, while requiring no human-labeled annotations. Genii consistently improves performance across different client agents during the polling, even when weaker models act as server agents. Further analysis reveals that Genii effectively mitigates judgment preference bias of LLM-based judgment models, demonstrating its effectiveness. All codes are available at https://github.com/NEUIR/Genii.

Xin Dai, Buqiang Xu, Zhenghao Liu et al.

Legal Artificial Intelligence (LegalAI) has achieved notable advances in automating judicial decision-making with the support of Large Language Models (LLMs). However, existing legal LLMs still struggle to generate reliable and interpretable reasoning processes. They often default to fast-thinking behavior by producing direct answers without explicit multi-step reasoning, limiting their effectiveness in complex legal scenarios that demand rigorous justification. To address this challenge, we propose Legal$Δ$, a reinforcement learning framework designed to enhance legal reasoning through chain-of-thought guided information gain. During training, Legal$Δ$ employs a dual-mode input setup-comprising direct answer and reasoning-augmented modes-and maximizes the information gain between them. This encourages the model to acquire meaningful reasoning patterns rather than generating superficial or redundant explanations. Legal$Δ$ follows a two-stage approach: (1) distilling latent reasoning capabilities from a powerful Large Reasoning Model (LRM), DeepSeek-R1, and (2) refining reasoning quality via differential comparisons, combined with a multidimensional reward mechanism that assesses both structural coherence and legal-domain specificity. Experimental results on multiple legal reasoning tasks demonstrate that Legal$Δ$ outperforms strong baselines in both accuracy and interpretability. It consistently produces more robust and trustworthy legal judgments without relying on labeled preference data. All code and data will be released at https://github.com/NEUIR/LegalDelta.

Buqiang Xu, Xin Dai, Zhenghao Liu et al.

Legal Judgment Prediction (LJP) is a fundamental task of legal artificial intelligence, aiming to automatically predict the judgment outcomes of legal cases. Existing LJP models primarily focus on identifying legal triggers within criminal fact descriptions by contrastively training language models. However, these LJP models overlook the importance of learning to effectively distinguish subtle differences among judgments, which is crucial for producing more accurate predictions. In this paper, we propose LegalDuet, which continuously pretrains language models to learn a more tailored embedding space for representing legal cases. Specifically, LegalDuet designs a dual-view mechanism to continuously pretrain language models: 1) Law Case Clustering retrieves similar cases as hard negatives and employs contrastive training to differentiate among confusing cases; 2) Legal Decision Matching aims to identify legal clues within criminal fact descriptions to align them with the chain of reasoning that contains the correct legal decision. Our experiments on the CAIL2018 dataset demonstrate the effectiveness of LegalDuet. Further analysis reveals that LegalDuet improves the ability of pretrained language models to distinguish confusing criminal charges by reducing prediction uncertainty and enhancing the separability of criminal charges. The experiments demonstrate that LegalDuet produces a more concentrated and distinguishable embedding space, effectively aligning criminal facts with corresponding legal decisions. The code is available at https://github.com/NEUIR/LegalDuet.

Yida Cai, Ranjuexiao Hu, Huiyuan Xie et al.

Legal relations form a highly consequential analytical framework of civil law system, serving as a crucial foundation for resolving disputes and realizing values of the rule of law in judicial practice. However, legal relations in Chinese civil cases remain underexplored in the field of legal artificial intelligence (legal AI), largely due to the absence of comprehensive schemas. In this work, we firstly introduce a comprehensive schema, which contains a hierarchical taxonomy and definitions of arguments, for AI systems to capture legal relations in Chinese civil cases. Based on this schema, we then formulate legal relation extraction task and present LexRel, an expert-annotated benchmark for legal relation extraction in Chinese civil law. We use LexRel to evaluate state-of-the-art large language models (LLMs) on legal relation extractions, showing that current LLMs exhibit significant limitations in accurately identifying civil legal relations. Furthermore, we demonstrate that incorporating legal relations information leads to consistent performance gains on other downstream legal AI tasks.

Xing Wang, Huiyuan Xie, Yiyan Wang et al.

Large language models (LLMs) are now deployed at unprecedented scale, assisting millions of users in daily tasks. However, the risk of these models assisting unlawful activities remains underexplored. In this study, we define this high-risk behavior as complicit facilitation - the provision of guidance or support that enables illicit user instructions - and present four empirical studies that assess its prevalence in widely deployed LLMs. Using real-world legal cases and established legal frameworks, we construct an evaluation benchmark spanning 269 illicit scenarios and 50 illicit intents to assess LLMs' complicit facilitation behavior. Our findings reveal widespread LLM susceptibility to complicit facilitation, with GPT-4o providing illicit assistance in nearly half of tested cases. Moreover, LLMs exhibit deficient performance in delivering credible legal warnings and positive guidance. Further analysis uncovers substantial safety variation across socio-legal contexts. On the legal side, we observe heightened complicity for crimes against societal interests, non-extreme but frequently occurring violations, and malicious intents driven by subjective motives or deceptive justifications. On the social side, we identify demographic disparities that reveal concerning complicit patterns towards marginalized and disadvantaged groups, with older adults, racial minorities, and individuals in lower-prestige occupations disproportionately more likely to receive unlawful guidance. Analysis of model reasoning traces suggests that model-perceived stereotypes, characterized along warmth and competence, are associated with the model's complicit behavior. Finally, we demonstrate that existing safety alignment strategies are insufficient and may even exacerbate complicit behavior.

Huiyuan Xie, Chenyang Li, Huining Zhu et al.

Legal reasoning is a fundamental component of legal analysis and decision-making. Existing computational approaches to legal reasoning predominantly rely on generic reasoning frameworks such as syllogism and IRAC, which do not comprehensively examine the nuanced processes that underpin legal reasoning. Moreover, current research has largely focused on criminal cases, with insufficient modeling for civil cases. In this work, we present a novel framework for explicitly modeling legal reasoning in the analysis of Chinese tort-related civil cases. We first operationalize the legal reasoning processes used in tort analysis into the LawChain framework. LawChain is a three-module reasoning framework, with each module consisting of multiple finer-grained sub-steps. Informed by the LawChain framework, we introduce the task of tort legal reasoning and construct an evaluation benchmark, LawChain$_{eval}$, to systematically assess the critical steps within analytical reasoning chains for tort analysis. Leveraging this benchmark, we evaluate state-of-the-art large language models for their legal reasoning ability in civil tort contexts. Our results indicate that current models still fall short in accurately handling crucial elements of tort legal reasoning. Furthermore, we introduce several baseline approaches that explicitly incorporate LawChain-style reasoning through prompting or post-training. We conduct further experiments on additional legal analysis tasks, such as Legal Named-Entity Recognition and Criminal Damages Calculation, to verify the generalizability of these baselines. The proposed baseline approaches achieve significant improvements in tort-related legal reasoning and generalize well to related legal analysis tasks, thus demonstrating the value of explicitly modeling legal reasoning chains to enhance the reasoning capabilities of language models.

Joana Ribeiro de Faria, Huiyuan Xie, Felix Steffek

Court transcripts and judgments are rich repositories of legal knowledge, detailing the intricacies of cases and the rationale behind judicial decisions. The extraction of key information from these documents provides a concise overview of a case, crucial for both legal experts and the public. With the advent of large language models (LLMs), automatic information extraction has become increasingly feasible and efficient. This paper presents a comprehensive study on the application of GPT-4, a large language model, for automatic information extraction from UK Employment Tribunal (UKET) cases. We meticulously evaluated GPT-4's performance in extracting critical information with a manual verification process to ensure the accuracy and relevance of the extracted data. Our research is structured around two primary extraction tasks: the first involves a general extraction of eight key aspects that hold significance for both legal specialists and the general public, including the facts of the case, the claims made, references to legal statutes, references to precedents, general case outcomes and corresponding labels, detailed order and remedies and reasons for the decision. The second task is more focused, aimed at analysing three of those extracted features, namely facts, claims and outcomes, in order to facilitate the development of a tool capable of predicting the outcome of employment law disputes. Through our analysis, we demonstrate that LLMs like GPT-4 can obtain high accuracy in legal information extraction, highlighting the potential of LLMs in revolutionising the way legal information is processed and utilised, offering significant implications for legal research and practice.

Huiyuan Xie, Zhenghao Liu, Chenyan Xiong et al.

Human conversations naturally evolve around different topics and fluently move between them. In research on dialog systems, the ability to actively and smoothly transition to new topics is often ignored. In this paper we introduce TIAGE, a new topic-shift aware dialog benchmark constructed utilizing human annotations on topic shifts. Based on TIAGE, we introduce three tasks to investigate different scenarios of topic-shift modeling in dialog settings: topic-shift detection, topic-shift triggered response generation and topic-aware dialog generation. Experiments on these tasks show that the topic-shift signals in TIAGE are useful for topic-shift response generation. On the other hand, dialog systems still struggle to decide when to change topic. This indicates further research is needed in topic-shift aware dialog modeling.

Huiyuan Xie, Tom Sherborne, Alexander Kuhnle et al.

Image captioning as a multimodal task has drawn much interest in recent years. However, evaluation for this task remains a challenging problem. Existing evaluation metrics focus on surface similarity between a candidate caption and a set of reference captions, and do not check the actual relation between a caption and the underlying visual content. We introduce a new diagnostic evaluation framework for the task of image captioning, with the goal of directly assessing models for grammaticality, truthfulness and diversity (GTD) of generated captions. We demonstrate the potential of our evaluation framework by evaluating existing image captioning models on a wide ranging set of synthetic datasets that we construct for diagnostic evaluation. We empirically show how the GTD evaluation framework, in combination with diagnostic datasets, can provide insights into model capabilities and limitations to supplement standard evaluations.

Alexander Kuhnle, Huiyuan Xie, Ann Copestake

The FiLM model achieves close-to-perfect performance on the diagnostic CLEVR dataset and is distinguished from other such models by having a comparatively simple and easily transferable architecture. In this paper, we investigate in more detail the ability of FiLM to learn various linguistic constructions. Our main results show that (a) FiLM is not able to learn relational statements straight away except for very simple instances, (b) training on a broader set of instances as well as pretraining on simpler instance types can help alleviate these learning difficulties, (c) mixing is less robust than pretraining and very sensitive to the compositional structure of the dataset. Overall, our results suggest that the approach of big all-encompassing datasets and the paradigm of "the effectiveness of data" may have fundamental limitations.