Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 2, 2026Proceedings of the ACM on software engineering.Open Access

Debugging Engine Enhanced by Prior Knowledge: Can We Teach LLM How to Debug?

View Full Paper
Ask AI
Bookmark
Share

Authors

KLKunyi LiSWSai WuXTXiu Tang

Discussion

Loading...

Member takes

Overview

Randomized trial evaluates DeepK for improving debugging knowledge and accuracy in program repair using LLMs, indicating significant advancements.

Key Points

  • The aim is to enhance automated program repair by effectively utilizing debugging knowledge in Large Language Models.
  • Introduced DeepK framework for systematic extraction, validation, and reuse of debugging knowledge.
  • Evaluated DeepK across multiple benchmarks (ACPR, Atcoder) using GPT-4o and DeepSeek-v3.
  • Conducted ablation studies to determine the impact of edit description generation and multi-perspective retrieval.
  • DeepK consistently surpassed state-of-the-art APR systems in repair accuracy.
  • Ablation studies confirmed the critical role of edit description generation and two-fold debugging knowledge.
  • Found a trade-off between informativeness and noise based on the number of retrieved entries.

Cite This Study

Li et al. (2026) studied this question.

synapsesocial.com/papers/6a45ffa29ed134303130ffb6https://doi.org/10.1145/3797110
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Exploring the Potential of Pre-Trained Language Models of Code for Automated Program Repair2024 · 3 citations
  2. 2Practical LLM-Based Function-Level Automated Program Repair: How Far Are We?2026 · 1 citations
  3. 3Self-bootstrapping automated program repair: using LLMs to generate and evaluate synthetic training data for bug repair2026
  4. 4A Systematic Literature Review on Large Language Models for Automated Program Repair2024 · 16 citations
  5. 5Hybrid Automated Program Repair by Combining Large Language Models and Program Analysis2024