CausalRepair: Bridging the Causality Gap in Large Language Model-Based Automated Program Repair via Dual-Slicing
Source
arxiv.org
Author
Linhao Wu, Yizhou Chen, Zhen Yang, Pengyu Xue, Dan Hao
Date
Context
CausalRepair tries to improve automated bug fixing by narrowing the evidence shown to a coding model. It analyzes test code to select relevant portions, then uses an execution trace to identify source-code dependencies involved in the failure. Together, these selections give the model a compact account of what the test expected and what the program actually executed. Research abstract.
Using DeepSeek-V3, the authors report 313 correct fixes on Defects4J. The contribution is a method for selecting repair context windowThe maximum amount of text a model can consider at once — its working memory for the current conversation or task.Full definition → around a particular failure, rather than requiring the model to identify the relevant evidence inside a larger code dump.
Terms in this piece · Glossary
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
articleRethinking Automated Program Repair: The Impact of Bug Complexity, Fault Localization, and LLM Cost-efficiencyJunchi Liu, Ali Bigdeli, Roya Daneshi, Atu Ambala, Sudipto Ghosh, Fabio Santos
articleMultiFixer: A Coordinator-Proposer Based Multi-Agent Framework For Fixing Multi-Hunk BugsHaichuan Hu, Chunrong Fang, Ye Shang, Jiawei Liu, Weifeng Sun, Guoqing Xie, Chenxing Zhong, Quanjun Zhang
articleHype Meets Reality: Large Language Models as Mutators in Search-based Automated Program Repair of Simulink-Stateflow ModelsAyesha Irshad, Pablo Valle, Jon Ayerdi, Aitor Arrieta