Lin Shi

Fang Liu, Yang Liu, Lin Shi et al.

The rise of Large Language Models (LLMs) has significantly advanced many applications on software engineering tasks, particularly in code generation. Despite the promising performance, LLMs are prone to generate hallucinations, which means LLMs might produce outputs that deviate from users' intent, exhibit internal inconsistencies, or misalign with the factual knowledge, making the deployment of LLMs potentially risky in a wide range of applications. Existing work mainly focuses on investing the hallucination in the domain of natural language generation (NLG), leaving a gap in understanding the types and extent of hallucinations in the context of code generation. To bridge the gap, we conducted a thematic analysis of the LLM-generated code to summarize and categorize the hallucinations present in it. Our study established a comprehensive taxonomy of hallucinations in LLM-generated code, encompassing 5 primary categories of hallucinations depending on the conflicting objectives and varying degrees of deviation observed in code generation. Furthermore, we systematically analyzed the distribution of hallucinations, exploring variations among different LLMs and their correlation with code correctness. Based on the results, we proposed HalluCode, a benchmark for evaluating the performance of code LLMs in recognizing hallucinations. Hallucination recognition and mitigation experiments with HalluCode and HumanEval show existing LLMs face great challenges in recognizing hallucinations, particularly in identifying their types, and are hardly able to mitigate hallucinations. We believe our findings will shed light on future research about hallucination evaluation, detection, and mitigation, ultimately paving the way for building more effective and reliable code LLMs in the future.

Chiyu Ma, Enpei Zhang, Yilun Zhao et al.

LLM-as-Judge has emerged as a scalable alternative to human evaluation, enabling large language models (LLMs) to provide reward signals in trainings. While recent work has explored multi-agent extensions such as multi-agent debate and meta-judging to enhance evaluation quality, the question of how intrinsic biases manifest in these settings remains underexplored. In this study, we conduct a systematic analysis of four diverse bias types: position bias, verbosity bias, chain-of-thought bias, and bandwagon bias. We evaluate these biases across two widely adopted multi-agent LLM-as-Judge frameworks: Multi-Agent-Debate and LLM-as-Meta-Judge. Our results show that debate framework amplifies biases sharply after the initial debate, and this increased bias is sustained in subsequent rounds, while meta-judge approaches exhibit greater resistance. We further investigate the incorporation of PINE, a leading single-agent debiasing method, as a bias-free agent within these systems. The results reveal that this bias-free agent effectively reduces biases in debate settings but provides less benefit in meta-judge scenarios. Our work provides a comprehensive study of bias behavior in multi-agent LLM-as-Judge systems and highlights the need for targeted bias mitigation strategies in collaborative evaluation settings.

Lin Shi, Chiyu Ma, Wenhua Liang et al.

LLM-as-a-Judge has emerged as a promising alternative to human evaluators across various tasks, yet inherent biases - particularly position bias, the tendency to favor solutions based on their position within the prompt - compromise its reliability. This exploratory study evaluates position bias in LLM judges across pairwise and list-wise comparison settings, introducing three metrics: repetition stability, position consistency, and preference fairness. Our experiments, involving 15 LLM judges across MTBench and DevBench with 22 tasks and approximately 40 solution-generating models, result in over 150,000 evaluation instances. We identify Judge-Level, Candidate-Level, and Task-Level factors contributing to bias. The findings confirm that position bias is not due to random chance and varies significantly across judges and tasks. While position bias is weakly influenced by the length of prompt components, it is strongly affected by the quality gap between solutions. Our agreement and disagreement analysis among judges further provides insights into the distribution of judging difficulty across the dataset, and highlights the potential for dataset modifications.