Lin Shi

Fangwen Mu, Junjie Wang, Lin Shi et al.

Automatically repairing software issues remains a fundamental challenge at the intersection of software engineering and AI. Although recent advances in Large Language Models (LLMs) have demonstrated potential for repository-level repair tasks, current methods exhibit two notable limitations: (1) they often address issues in isolation, neglecting to incorporate insights from previously resolved issues, and (2) they rely on static, rigid prompting strategies that constrain their ability to generalize across diverse and evolving contexts. We propose ExpeRepair, a novel LLM-based program repair framework inspired by the dual-memory systems of human cognition, where episodic and semantic memory synergistically support learning and decision-making. Unlike existing methods, ExpeRepair continuously learns from historical repair experiences via dual-channel knowledge accumulation, enabling it to adaptively reuse past knowledge during inference. Specifically, ExpeRepair organizes prior repair knowledge into two complementary memories: an episodic memory that stores concrete repair demonstrations, and a semantic memory that encodes abstract, reflective insights. At inference time, ExpeRepair activates both memory systems by retrieving relevant demonstrations from episodic memory and recalling high-level repair insights from semantic memory. It further enhances adaptability through dynamic prompt composition, integrating both memory types to replace static prompts with context-aware, experience-driven prompts. We evaluate ExpeRepair on two benchmarks: SWE-Bench Lite and SWE-Bench Verified. Experimental results show that ExpeRepair achieves pass@1 scores of 60.3% and 74.6% on the two benchmarks, respectively, achieving the best performance among the evaluated open-source methods. We have open-sourced ExpeRepair at https://github.com/ExpeRepair/ExpeRepair.

Yuhong Liu, Yunhe Su, Zhipeng Peng et al.

With the advent of large language models, research in automated software engineering has increasingly focused on leveraging these models to achieve a deeper semantic understanding of code or to engineer sophisticated agent-based processes. However, this research trajectory has largely overlooked a critical factor: the developers themselves. Programming is a deeply individualized activity; developers exhibit significant variation in their tool-chain preferences, domain-specific expertise, and problem-solving strategies. Consequently, the current paradigm of one-size-fits-all code intelligence systems struggles to accommodate the needs of individual developers. To address this gap, we introduce VirtualME, a novel IDE-embedded data infrastructure designed to model the developer by continuously capturing and interpreting their dynamic programming behaviors and preferences. VirtualME contains three components. (1) Log-level Behavior Extraction: it captures and extracts developers' log-level behaviors from IDE. (2) Task-level Behavior Recognition: it aggregates log-level behaviors into task-level behaviors via a multi-agent pipeline. (3) Developer-personality Measurement: it builds a rule engine to distill a four-dimensional developer persona: technology stack, ability, behavioral habits, and learning style. On top of VirtualME, we propose a solution for personalized repository-level knowledge Q&A by integrating the developer persona into the Q&A agent. We evaluated VirtualME by building a multi-repository benchmark with real-world developer trajectories, balancing correctness and personalization. Experimental results show that VirtualME-enhanced answers outperform generic baselines on five dimensions, yielding an average 33.80% improvement. Our results demonstrate that abundant, continuous developer-behavior data can pave the new way for adaptive and personalized code intelligence.

Fangwen Mu, Xiao Chen, Lin Shi et al.

Deliberation is a common and natural behavior in human daily life. For example, when writing papers or articles, we usually first write drafts, and then iteratively polish them until satisfied. In light of such a human cognitive process, we propose DECOM, which is a multi-pass deliberation framework for automatic comment generation. DECOM consists of multiple Deliberation Models and one Evaluation Model. Given a code snippet, we first extract keywords from the code and retrieve a similar code fragment from a pre-defined corpus. Then, we treat the comment of the retrieved code as the initial draft and input it with the code and keywords into DECOM to start the iterative deliberation process. At each deliberation, the deliberation model polishes the draft and generates a new comment. The evaluation model measures the quality of the newly generated comment to determine whether to end the iterative process or not. When the iterative process is terminated, the best-generated comment will be selected as the target comment. Our approach is evaluated on two real-world datasets in Java (87K) and Python (108K), and experiment results show that our approach outperforms the state-of-the-art baselines. A human evaluation study also confirms the comments generated by DECOM tend to be more readable, informative, and useful.

Lin Shi, Qiyuan Liu, Bei Peng

Relative overgeneralization (RO) is a pathology that can arise in cooperative multi-agent tasks when the optimal joint action's utility falls below that of a sub-optimal joint action. RO can cause the agents to get stuck into local optima or fail to solve cooperative tasks requiring significant coordination between agents within a given timestep. In this work, we empirically find that, in multi-agent reinforcement learning (MARL), both value-based and policy gradient MARL algorithms can suffer from RO and fail to learn effective coordination policies. To better overcome RO, we propose a novel approach called curriculum learning for relative overgeneralization (CURO). To solve a target task that exhibits strong RO, in CURO, we first fine-tune the reward function of the target task to generate source tasks to train the agent. Then, to effectively transfer the knowledge acquired in one task to the next, we use a transfer learning method that combines value function transfer with buffer transfer, which enables more efficient exploration in the target task. CURO is general and can be applied to both value-based and policy gradient MARL methods. We demonstrate that, when applied to QMIX, HAPPO, and HATRPO, CURO can successfully overcome severe RO, achieve improved performance, and outperform baseline methods in a variety of challenging cooperative multi-agent tasks.

Liyou Chen, Hailong Sun, Xiang Gao et al.

The utilization of third-party open-source libraries is widespread in modern software development. Due to the dependency relationships, vulnerabilities within open-source libraries pose significant security threats to downstream software. However, the library vulnerabilities are usually implicitly reported and patched, without explicit notification to dependent software, leaving the downstream software vulnerable to potential attacks. Existing research efforts primarily focus on identifying vulnerability patches according to bug reports, commit messages, or code changes, overlooking the rich semantic connections among various sources of information. In this paper, our main insight is that various sources of information, including the vulnerability descriptions (e.g., bug reports) and its fixing strategies (e.g., commit messages and code changes), are highly interconnected. They express the high-level semantic information about the symptom, root cause and fixing strategies of the bugs. Hence, we propose an approach that involves training an AI model to integrate multiple sources, thus enhancing the effectiveness of vulnerability identification and vulnerability type classification. We introduce VPFinder, a tool that utilizes multi-head attention mechanisms to extract high-level semantic information from diverse sources. Evaluation results demonstrate that VPFinder achieves remarkable 0.941 F1-score in vulnerability identification task and 0.610 F1-score in vulnerability type classification task, outperforming state-of-the-art approaches by 5.4%.

Yurong Wu, Fangwen Mu, Qiuhong Zhang et al.

Prompt trading has emerged as a significant intellectual property concern in recent years, where vendors entice users by showcasing sample images before selling prompt templates that can generate similar images. This work investigates a critical security vulnerability: attackers can steal prompt templates using only a limited number of sample images. To investigate this threat, we introduce Prism, a prompt-stealing benchmark consisting of 50 templates and 450 images, organized into Easy and Hard difficulty levels. To identify the vulnerabity of VLMs to prompt stealing, we propose EvoStealer, a novel template stealing method that operates without model fine-tuning by leveraging differential evolution algorithms. The system first initializes population sets using multimodal large language models (MLLMs) based on predefined patterns, then iteratively generates enhanced offspring through MLLMs. During evolution, EvoStealer identifies common features across offspring to derive generalized templates. Our comprehensive evaluation conducted across open-source (INTERNVL2-26B) and closed-source models (GPT-4o and GPT-4o-mini) demonstrates that EvoStealer's stolen templates can reproduce images highly similar to originals and effectively generalize to other subjects, significantly outperforming baseline methods with an average improvement of over 10%. Moreover, our cost analysis reveals that EvoStealer achieves template stealing with negligible computational expenses. Our code and dataset are available at https://github.com/whitepagewu/evostealer.

Binquan Zhang, Li Zhang, Lin Shi et al.

Large Language Models (LLMs) have revolutionized code generation, evolving from static tools into dynamic conversational interfaces that facilitate complex, multi-turn collaborative programming. While LLMs exhibit remarkable proficiency in generating standalone code snippets, they often struggle to maintain contextual consistency during extended interactions, creating significant obstacles in the collaboration process. Existing benchmarks primarily emphasize the functional correctness of the final output, overlooking latent quality issues within the interaction process itself, which we term Interaction Smells. In this paper, we conduct an empirical study on sampled real-word user-LLM interactions from WildChat and LMSYS-Chat-1M datasets to systematically investigate Interaction Smells in human-LLM code generation tasks from the perspectives of phenomena, distribution, and mitigation. First, we establish the first taxonomy of Interaction Smells by manually performing open card sorting on real-world interaction logs. This taxonomy categorizes Interaction Smells into three primary categories, i.e., User Intent Quality, Historical Instruction Compliance, and Historical Response Violation, comprising nine specific subcategories. Next, we quantitatively evaluate six mainstream LLMs (i.e., GPT-4o, DeepSeek-Chat, Gemini 2.5, Qwen2.5-32B, Qwen2.5-72B, and Qwen3-235B-a22b) to analyze the distribution of Interaction Smells across different models. Finally, we propose Invariant-aware Constraint Evolution (InCE), a multi-agent framework designed to improve multi-turn interaction quality through explicit extraction of global invariants and pre-generation quality audits. Experimental results on the extended WildBench benchmark demonstrate that this lightweight mitigation approach significantly improves the Task Success Rate and effectively suppresses the occurrence of Interaction Smells.

Ziyou Jiang, Lin Shi, Guowei Yang et al.

Security patches are essential for enhancing the stability and robustness of projects in the software community. While vulnerabilities are officially expected to be patched before being disclosed, patching vulnerabilities is complicated and remains a struggle for many organizations. To patch vulnerabilities, security practitioners typically track vulnerable issue reports (IRs), and analyze their relevant insecure code to generate potential patches. However, the relevant insecure code may not be explicitly specified and practitioners cannot track the insecure code in the repositories, thus limiting their ability to generate patches. In such cases, providing examples of insecure code and the corresponding patches would benefit the security developers to better locate and fix the insecure code. In this paper, we propose PatUntrack to automatically generating patch examples from IRs without tracked insecure code. It auto-prompts Large Language Models (LLMs) to make them applicable to analyze the vulnerabilities. It first generates the completed description of the Vulnerability-Triggering Path (VTP) from vulnerable IRs. Then, it corrects hallucinations in the VTP description with external golden knowledge. Finally, it generates Top-K pairs of Insecure Code and Patch Example based on the corrected VTP description. To evaluate the performance, we conducted experiments on 5,465 vulnerable IRs. The experimental results show that PatUntrack can obtain the highest performance and improve the traditional LLM baselines by +14.6% (Fix@10) on average in patch example generation. Furthermore, PatUntrack was applied to generate patch examples for 76 newly disclosed vulnerable IRs. 27 out of 37 replies from the authors of these IRs confirmed the usefulness of the patch examples generated by PatUntrack, indicating that they can benefit from these examples for patching the vulnerabilities.

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.

Lin Shi, Ziyou Jiang, Ye Yang et al.

Collaborative live chats are gaining popularity as a development communication tool. In community live chatting, developers are likely to post issues they encountered (e.g., setup issues and compile issues), and other developers respond with possible solutions. Therefore, community live chats contain rich sets of information for reported issues and their corresponding solutions, which can be quite useful for knowledge sharing and future reuse if extracted and restored in time. However, it remains challenging to accurately mine such knowledge due to the noisy nature of interleaved dialogs in live chat data. In this paper, we first formulate the problem of issue-solution pair extraction from developer live chat data, and propose an automated approach, named ISPY, based on natural language processing and deep learning techniques with customized enhancements, to address the problem. Specifically, ISPY automates three tasks: 1) Disentangle live chat logs, employing a feedforward neural network to disentangle a conversation history into separate dialogs automatically; 2) Detect dialogs discussing issues, using a novel convolutional neural network (CNN), which consists of a BERT-based utterance embedding layer, a context-aware dialog embedding layer, and an output layer; 3) Extract appropriate utterances and combine them as corresponding solutions, based on the same CNN structure but with different feeding inputs. To evaluate ISPY, we compare it with six baselines, utilizing a dataset with 750 dialogs including 171 issue-solution pairs and evaluate ISPY from eight open source communities. The results show that, for issue-detection, our approach achieves the F1 of 76%, and outperforms all baselines by 30%. Our approach achieves the F1 of 63% for solution-extraction and outperforms the baselines by 20%.

Lin Shi, Xiao Chen, Ye Yang et al.

Modern communication platforms such as Gitter and Slack play an increasingly critical role in supporting software teamwork, especially in open source development.Conversations on such platforms often contain intensive, valuable information that may be used for better understanding OSS developer communication and collaboration. However, little work has been done in this regard. To bridge the gap, this paper reports a first comprehensive empirical study on developers' live chat, investigating when they interact, what community structures look like, which topics are discussed, and how they interact. We manually analyze 749 dialogs in the first phase, followed by an automated analysis of over 173K dialogs in the second phase. We find that developers tend to converse more often on weekdays, especially on Wednesdays and Thursdays (UTC), that there are three common community structures observed, that developers tend to discuss topics such as API usages and errors, and that six dialog interaction patterns are identified in the live chat communities. Based on the findings, we provide recommendations for individual developers and OSS communities, highlight desired features for platform vendors, and shed light on future research directions. We believe that the findings and insights will enable a better understanding of developers' live chat, pave the way for other researchers, as well as a better utilization and mining of knowledge embedded in the massive chat history.

Chenyang Zhu, Bin Xiao, Lin Shi et al.

The recent Segment Anything Model (SAM) represents a significant breakthrough in scaling segmentation models, delivering strong performance across various downstream applications in the RGB modality. However, directly applying SAM to emerging visual modalities, such as depth and event data results in suboptimal performance in multi-modal segmentation tasks. In this paper, we make the first attempt to adapt SAM for multi-modal semantic segmentation by proposing a Mixture of Low-Rank Adaptation Experts (MoE-LoRA) tailored for different input visual modalities. By training only the MoE-LoRA layers while keeping SAM's weights frozen, SAM's strong generalization and segmentation capabilities can be preserved for downstream tasks. Specifically, to address cross-modal inconsistencies, we propose a novel MoE routing strategy that adaptively generates weighted features across modalities, enhancing multi-modal feature integration. Additionally, we incorporate multi-scale feature extraction and fusion by adapting SAM's segmentation head and introducing an auxiliary segmentation head to combine multi-scale features for improved segmentation performance effectively. Extensive experiments were conducted on three multi-modal benchmarks: DELIVER, MUSES, and MCubeS. The results consistently demonstrate that the proposed method significantly outperforms state-of-the-art approaches across diverse scenarios. Notably, under the particularly challenging condition of missing modalities, our approach exhibits a substantial performance gain, achieving an improvement of 32.15% compared to existing methods.

Bowen Li, Wenhan Wu, Ziwei Tang et al.

Recent advancements in large language models (LLMs) have significantly enhanced their coding capabilities. However, existing benchmarks predominantly focused on simplified or isolated aspects of coding, such as single-file code generation or repository issue debugging, falling short of measuring the full spectrum of challenges raised by real-world programming activities. In this case study, we explore the performance of LLMs across the entire software development lifecycle with DevEval, encompassing stages including software design, environment setup, implementation, acceptance testing, and unit testing. DevEval features four programming languages, multiple domains, high-quality data collection, and carefully designed and verified metrics for each task. Empirical studies show that current LLMs, including GPT-4, fail to solve the challenges presented within DevEval. Our findings offer actionable insights for the future development of LLMs toward real-world programming applications.

Peiding Wang, Li Zhang, Fang Liu et al.

Large Language Models (LLMs) have demonstrated exceptional performance in code generation tasks and have become indispensable programming assistants for developers. However, existing code generation benchmarks primarily assess the functional correctness of code generated by LLMs in single-turn interactions. They offer limited insight into LLMs' abilities to generate code that strictly follows users' instructions in multi-turn interaction scenarios. In this paper, we introduce CodeIF-Bench, a benchmark for evaluating the instruction-following capabilities of LLMs in interactive code generation. Specifically, CodeIF-Bench incorporates nine types of verifiable instructions aligned with the real-world software development requirements, which can be independently and objectively validated through specified test cases, facilitating the evaluation of instruction-following capability in multi-turn interactions. In both \textit{Static Conversation} and \textit{Dynamic Conversation} settings, we evaluate the performance of 7 state-of-the-art LLMs and summarize the important factors influencing the instruction-following ability of LLMs in multi-turn interactions, as well as potential directions for improvement.

Yutong Xie, Yiyao Liu, Zhuang Ma et al.

The deployment of large language models (LLMs) in diverse applications requires a thorough understanding of their decision-making strategies and behavioral patterns. As a supplement to a recent study on the behavioral Turing test, this paper presents a comprehensive analysis of five leading LLM-based chatbot families as they navigate a series of behavioral economics games. By benchmarking these AI chatbots, we aim to uncover and document both common and distinct behavioral patterns across a range of scenarios. The findings provide valuable insights into the strategic preferences of each LLM, highlighting potential implications for their deployment in critical decision-making roles.

Fangwen Mu, Junjie Wang, Zhuohao Yu et al.

Neural code models have found widespread success in tasks pertaining to code intelligence, yet they are vulnerable to backdoor attacks, where an adversary can manipulate the victim model's behavior by inserting triggers into the source code. Recent studies indicate that advanced backdoor attacks can achieve nearly 100% attack success rates on many software engineering tasks. However, effective defense techniques against such attacks remain insufficiently explored. In this study, we propose CodePurify, a novel defense against backdoor attacks on code models through entropy-based purification. Entropy-based purification involves the process of precisely detecting and eliminating the possible triggers in the source code while preserving its semantic information. Within this process, CodePurify first develops a confidence-driven entropy-based measurement to determine whether a code snippet is poisoned and, if so, locates the triggers. Subsequently, it purifies the code by substituting the triggers with benign tokens using a masked language model. We extensively evaluate CodePurify against four advanced backdoor attacks across three representative tasks and two popular code models. The results show that CodePurify significantly outperforms four commonly used defense baselines, improving average defense performance by at least 40%, 40%, and 12% across the three tasks, respectively. These findings highlight the potential of CodePurify to serve as a robust defense against backdoor attacks on neural code models.

Zexin Wang, Lin Shi, Haoyu Wu et al.

Epilepsy is a prevalent neurological disorder marked by sudden, brief episodes of excessive neuronal activity caused by abnormal electrical discharges, which may lead to some mental disorders. Most existing deep learning methods for epilepsy detection rely solely on unimodal EEG signals, neglecting the potential benefits of multimodal information. To address this, we propose a novel multimodal model, DistilCLIP-EEG, based on the CLIP framework, which integrates both EEG signals and text descriptions to capture comprehensive features of epileptic seizures. The model involves an EEG encoder based on the Conformer architecture as a text encoder, the proposed Learnable BERT (BERT-LP) as prompt learning within the encoders. Both operate in a shared latent space for effective cross-modal representation learning. To enhance efficiency and adaptability, we introduce a knowledge distillation method where the trained DistilCLIP-EEG serves as a teacher to guide a more compact student model to reduce training complexity and time. On the TUSZ, AUBMC, and CHB-MIT datasets, both the teacher and student models achieved accuracy rates exceeding 97%. Across all datasets, the F1-scores were consistently above 0.94, demonstrating the robustness and reliability of the proposed framework. Moreover, the student model's parameter count and model size are approximately 58.1% of those of the teacher model, significantly reducing model complexity and storage requirements while maintaining high performance. These results highlight the potential of our proposed model for EEG-based epilepsy detection and establish a solid foundation for deploying lightweight models in resource-constrained settings.

Jiayuan She, Lin Shi, Peiqi Li et al.

Background. Infectious diseases, particularly COVID-19, continue to be a significant global health issue. Although many countries have reduced or stopped large-scale testing measures, the detection of such diseases remains a propriety. Objective. This study aims to develop a novel, lightweight deep neural network for efficient, accurate, and cost-effective detection of COVID-19 using a nasal breathing audio data collected via smartphones. Methodology. Nasal breathing audio from 128 patients diagnosed with the Omicron variant was collected. Mel-Frequency Cepstral Coefficients (MFCCs), a widely used feature in speech and sound analysis, were employed for extracting important characteristics from the audio signals. Additional feature selection was performed using Random Forest (RF) and Principal Component Analysis (PCA) for dimensionality reduction. A Dense-ReLU-Dropout model was trained with K-fold cross-validation (K=3), and performance metrics like accuracy, precision, recall, and F1-score were used to evaluate the model. Results. The proposed model achieved 97% accuracy in detecting COVID-19 from nasal breathing sounds, outperforming state-of-the-art methods such as those by [23] and [13]. Our Dense-ReLU-Dropout model, using RF and PCA for feature selection, achieves high accuracy with greater computational efficiency compared to existing methods that require more complex models or larger datasets. Conclusion. The findings suggest that the proposed method holds significant potential for clinical implementation, advancing smartphone-based diagnostics in infectious diseases. The Dense-ReLU-Dropout model, combined with innovative feature processing techniques, offers a promising approach for efficient and accurate COVID-19 detection, showcasing the capabilities of mobile device-based diagnostics

Xingjian Diao, Zheyuan Liu, Chunhui Zhang et al.

Large Vision-Language Models (LVLMs) have exhibited strong reasoning capabilities through chain-of-thought mechanisms that generate step-by-step rationales. However, such slow-thinking approaches often lead to overthinking, where models produce excessively verbose responses even for simple queries, resulting in test-time inefficiency and even degraded accuracy. Prior work has attempted to mitigate this issue via adaptive reasoning strategies, but these methods largely overlook a fundamental bottleneck: visual perception failures. We argue that stable reasoning critically depends on low-level visual grounding, and that reasoning errors often originate from imperfect perception rather than insufficient deliberation. To address this limitation, we propose Gated Perception-Reasoning Optimization (GPRO), a meta-reasoning controller that dynamically routes computation among three decision paths at each generation step: a lightweight fast path, a slow perception path for re-examining visual inputs, and a slow reasoning path for internal self-reflection. To learn this distinction, we derive large-scale failure attribution supervision from approximately 790k samples, using teacher models to distinguish perceptual hallucinations from reasoning errors. We then train the controller with multi-objective reinforcement learning to optimize the trade-off between task accuracy and computational cost under uncertainty. Experiments on five benchmarks demonstrate that GPRO substantially improves both accuracy and efficiency, outperforming recent slow-thinking methods while generating significantly shorter responses.

Yachuan Liu, Xiaochun Wei, Lin Shi et al.

Large language models (LLMs) face significant challenges in ex-ante reasoning, where analysis, inference, or predictions must be made without access to information from future events. Even with explicit prompts enforcing temporal cutoffs, LLMs often generate outputs influenced by internalized knowledge of events beyond the specified cutoff. This paper introduces a novel task and benchmark designed to evaluate the ability of LLMs to reason while adhering to such temporal constraints. The benchmark includes a variety of tasks: stock prediction, Wikipedia event prediction, scientific publication prediction, and Question Answering (QA), designed to assess factual knowledge under temporal cutoff constraints. We use leakage rate to quantify models' reliance on future information beyond cutoff timestamps. Experimental results reveal that LLMs struggle to consistently adhere to temporal cutoffs across common prompting strategies and tasks, demonstrating persistent challenges in ex-ante reasoning. This benchmark provides a potential evaluation framework to advance the development of LLMs' temporal reasoning ability for time-sensitive applications.

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.

Qianhui Zhao, Li Zhang, Fang Liu et al.

Code generation is a latency-sensitive task that demands high timeliness. However, with the growing interest and inherent difficulty in repository-level code generation, most existing code generation studies focus on improving the correctness of generated code while overlooking the inference efficiency, which is substantially affected by the overhead during LLM generation. Although there has been work on accelerating LLM inference, these approaches are not tailored to the specific characteristics of code generation; instead, they treat code the same as natural language sequences and ignore its unique syntax and semantic characteristics, which are also crucial for improving efficiency. Consequently, these approaches exhibit limited effectiveness in code generation tasks, particularly for repository-level scenarios with considerable complexity and difficulty. To alleviate this issue, following draft-verification paradigm, we propose FastCoder, a simple yet highly efficient inference acceleration approach specifically designed for code generation, without compromising the quality of the output. FastCoder constructs a multi-source datastore, providing access to both general and project-specific knowledge, facilitating the retrieval of high-quality draft sequences. Moreover, FastCoder reduces the retrieval cost by controlling retrieval timing, and enhances efficiency through parallel retrieval and a context- and LLM preference-aware cache. Experimental results show that FastCoder can reach up to 2.53x and 2.54x speedup compared to autoregressive decoding in repository-level and standalone code generation tasks, respectively, outperforming state-of-the-art inference acceleration approaches by up to 88%. FastCoder can also be integrated with existing correctness-focused code generation approaches to accelerate the LLM generation process, and reach a speedup exceeding 2.6x.

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.

Shoubin Li, Xuyan Ma, Shuaiqun Pan et al.

Documents often contain complex physical structures, which make the Document Layout Analysis (DLA) task challenging. As a pre-processing step for content extraction, DLA has the potential to capture rich information in historical or scientific documents on a large scale. Although many deep-learning-based methods from computer vision have already achieved excellent performance in detecting \emph{Figure} from documents, they are still unsatisfactory in recognizing the \emph{List}, \emph{Table}, \emph{Text} and \emph{Title} category blocks in DLA. This paper proposes a VTLayout model fusing the documents' deep visual, shallow visual, and text features to localize and identify different category blocks. The model mainly includes two stages, and the three feature extractors are built in the second stage. In the first stage, the Cascade Mask R-CNN model is applied directly to localize all category blocks of the documents. In the second stage, the deep visual, shallow visual, and text features are extracted for fusion to identify the category blocks of documents. As a result, we strengthen the classification power of different category blocks based on the existing localization technique. The experimental results show that the identification capability of the VTLayout is superior to the most advanced method of DLA based on the PubLayNet dataset, and the F1 score is as high as 0.9599.

Ziyou Jiang, Lin Shi, Celia Chen et al.

Despite the valuable information contained in software chat messages, disentangling them into distinct conversations is an essential prerequisite for any in-depth analyses that utilize this information. To provide a better understanding of the current state-of-the-art, we evaluate five popular dialog disentanglement approaches on software-related chat. We find that existing approaches do not perform well on disentangling software-related dialogs that discuss technical and complex topics. Further investigation on how well the existing disentanglement measures reflect human satisfaction shows that existing measures cannot correctly indicate human satisfaction on disentanglement results. Therefore, in this paper, we introduce and evaluate a novel measure, named DLD. Using results of human satisfaction, we further summarize four most frequently appeared bad disentanglement cases on software-related chat to insight future improvements. These cases include (i) ignoring interaction patterns; (ii) ignoring contextual information; (iii) mixing up topics; and (iv) ignoring user relationships. We believe that our findings provide valuable insights on the effectiveness of existing dialog disentanglement approaches and these findings would promote a better application of dialog disentanglement in software engineering.

Anthony Hei-Long Chan, Yishan Luo, Lin Shi et al.

We propose a disease classification model, called the QC-SPHARM, for the early detection of the Alzheimer's Disease (AD). The proposed QC-SPHARM can distinguish between normal control (NC) subjects and AD patients, as well as between amnestic mild cognitive impairment (aMCI) patients having high possibility progressing into AD and those who do not. Using the spherical harmonics (SPHARM) based registration, hippocampal surfaces segmented from the ADNI data are individually registered to a template surface constructed from the NC subjects using SPHARM. Local geometric distortions of the deformation from the template surface to each subject are quantified in terms of conformality distortions and curvatures distortions. The measurements are combined with the spherical harmonics coefficients and the total volume change of the subject from the template. Afterwards, a t-test based feature selection method incorporating the bagging strategy is applied to extract those local regions having high discriminating power of the two classes. The disease diagnosis machine can therefore be built using the data under the Support Vector Machine (SVM) setting. Using 110 NC subjects and 110 AD patients from the ADNI database, the proposed algorithm achieves 85:2% testing accuracy on 80 random samples as testing subjects, with the incorporation of surface geometry in the classification machine. Using 20 aMCI patients who has advanced to AD during a two-year period and another 20 aMCI patients who remain non-AD for the next two years, the algorithm achieves 81:2% accuracy using 10 randomly picked subjects as testing data. Our proposed method is 6%-15% better than other classification models without the incorporation of surface geometry. The results demonstrate the advantages of using local geometric distortions as the discriminating criterion for early AD diagnosis.