ISSTA 是 APR 的主战场

Face It Yourselves: An LLM-Based Two-Stage Strategy to Localize Configuration Errors via Logs

Detecting Build Dependency Errors in Incremental Builds

FastLog: An End-to-End Method to Efficiently Generate and Insert Logging Statements

CEBin: A Cost-Effective Framework for Large-Scale Binary Code Similarity Detection

Model-less Is the Best Model: Generating Pure Code Implementations to Replace On-Device DL Models

LPR: Large Language Models-Aided Program Reduction

Bridge and Hint: Extending Pre-trained Language Models for Long-Range Code

Evaluating the Effectiveness of Decompilers

CLAP: Learning Transferable Binary Code Representations with Natural Language Supervision

Automated Program Repair via Conversation: Fixing 162 out of 337 Bugs for $0.42 Each using ChatGPT

背景：

  • 传统 APR 通常使用生成——验证范式。但是是生成一堆然后让人类检查，极易导致重复和反复错误
  • 基于大模型的 APR 受限于训练数据，并且解决不了没见过的场景
  • 局限性：
    • 没充分考虑报错信息
    • 模型会产生大量重复样本，并且和人类修正 bug 的方式不一样
    • LLM没有利用之前的正确修正作为后续的补丁学习和参考
  • 作者提出的 ChatAPR 是完全自动化的对话驱动的 APR
    image-20250710190652.webp

点评：当今水论文的趋势，借助大模型的威力来 Prompt Engineering。论文背景介绍的比较全面，可以学习

CREF: An LLM-Based Conversational Software Repair Framework for Programming Tutors

AI Coders Are among Us: Rethinking Programming Language Grammar towards Efficient Code Generation

SelfPiCo: Self-Guided Partial Code Execution with LLMs

One Size Does Not Fit All: Multi-granularity Patch Generation for Better Automated Program Repair

Panda: A Concurrent Scheduler for Compiler-Based Tools