Issue Localization via LLM-Driven Iterative Code Graph Searching
This program is tentative and subject to change.
Issue solving aims to generate patches to fix reported issues in real-world code repositories according to issue descriptions. Issue localization forms the basis for accurate issue solving. Recently, large language model (LLM) based issue localization methods have demonstrated state-of-the-art performance. However, these methods either search from files mentioned in issue descriptions or in the whole repository and struggle to balance the breadth and depth of the search space to converge on the target efficiently. Moreover, they allow LLM to explore whole repositories freely, making it challenging to control the search direction to prevent the LLM from searching for incorrect targets. Meanwhile, because LLMs may not correctly produce the required interaction formats with the environment, they suffer from search failures.
This paper introduces IGSIL, an LLM-driven, powerful function-level issue localization method without training or indexing. To balance search breadth and depth, IGSIL employs a two-phase code graph search strategy. It first conducts broad exploration at the file level using dynamically constructed module call graphs, and then performs in-depth analysis at the function level by expanding the module call graph into a function call graph and executing iterative searches. To precisely control the search direction, IGSIL designs a pruner to filter unrelated directions and irrelevant contexts. To avoid incorrect interaction formats in long contexts, IGSIL introduces a reflection mechanism that uses additional independent queries in short contexts to enhance formatted abilities. Experiment results demonstrate that IGSIL achieves a Top-1 localization accuracy of 43.3% and 44.6% on SWE-bench Lite and SWE-bench Verified, respectively, with Qwen2.5-Coder-32B, average outperforming the state-of-the-art methods by 96.04%. When IGSIL is integrated into an issue-solving method, Agentless, the issue resolution rate improves by 2.98%–30.5%.
This program is tentative and subject to change.
Tue 18 NovDisplayed time zone: Seoul change
11:00 - 12:30 | |||
11:00 10mTalk | FaultSeeker: LLM-Empowered Framework for Blockchain Transaction Fault Localization Research Papers Kairan Sun Nanyang Technological University, Zhengzi Xu Imperial Global Singapore, Kaixuan Li Nanyang Technological University, Lyuye Zhang Nanyang Technological University, Yuqiang Sun Nanyang Technological University, Liwei Tan MetaTrust Labs, Yang Liu Nanyang Technological University | ||
11:10 10mTalk | FlexFL: Flexible and Effective Fault Localization With Open-Source Large Language Models Journal-First Track Chuyang Xu Zhejiang University, Zhongxin Liu Zhejiang University, Xiaoxue Ren Zhejiang University, Gehao Zhang Ant Group, Ming Liang Ant Group, David Lo Singapore Management University | ||
11:20 10mTalk | LLM-Based Identification of Null Pointer Exception Patches Research Papers Tahir Ullah Beijing Institute of Technology, Waseem Akram Beijing Institute of Technology, Fiza Khaliq Beijing Institute of Technology, Hui Liu Beijing Institute of Technology | ||
11:30 10mTalk | SpectAcle: Fault Localisation of AI-Enabled CPS by Exploiting Sequences of DNN Controller Inferences Journal-First Track Deyun Lyu National Institute of Informatics, Zhenya Zhang Kyushu University, Japan, Paolo Arcaini National Institute of Informatics
, Xiao-Yi Zhang University of Science and Technology Beijing, Fuyuki Ishikawa National Institute of Informatics, Jianjun Zhao Kyushu University | ||
11:40 10mTalk | Sifting Truth from Coincidences: A Two-Stage Positive and Unlabeled Learning Model for Coincidental Correctness Detection Research Papers Chunyan Liu Chongqing University, Huan Xie Chongqing University, Yan Lei Chongqing University, Zhenyu Wu School of Big Data & Software Engineering, Chongqing University, Jinping Wang Chonqing University | ||
11:50 10mTalk | Let the Code Speak: Incorporating Program Dynamic State for Better Method-Level Fault Localization Research Papers Yihao Qin , Shangwen Wang National University of Defense Technology, Bo Lin National University of Defense Technology, Xin Peng , Sheng Ouyang National University of Defense Technology, Liqian Chen National University of Defense Technology, Xiaoguang Mao National University of Defense Technology | ||
12:00 10mTalk | Issue Localization via LLM-Driven Iterative Code Graph Searching Research Papers Zhonghao Jiang Zhejiang University, Xiaoxue Ren Zhejiang University, Meng Yan Chongqing University, Wei Jiang Ant Group, Yong Li Ant Group, Zhongxin Liu Zhejiang University | ||
12:10 10mTalk | Hypergraph Neural Network-based Multi-Granular Root Cause Localization for Microservice Systems Research Papers Yaxiao Li Xidian University, Lu Wang Xidian University, Chenxi Zhang Xidian University, Qingshan Li Xidian University, Siming Rong Xidian University, Baiyang Wen Xidian University, Xuyang Li Purdue University, Kun Ma Xidian University, Quanwei Du Xidian University, KeYang Li Xidian University, Lingfeng Pan Xidian University, Xinyue Li Peking University, MingXuan Hui Xidian University | ||
12:20 10mTalk | Explainable Fault Localization for Programming Assignments via LLM-Guided Annotation Research Papers Fang Liu Beihang University, Tianze Wang Beihang University, Li Zhang Beihang University, Zheyu Yang Beihang University, Jing Jiang Beihang University, Zian Sun Beihang University Pre-print | ||