Adnan KhanAdnanProduct & AI
San Francisco Bay Area
Get in touch

Source note3 min read

Multi-Perspective Agentic Program Repair via Code Property Graphs and Temporal Execution Graphs

Source

arXiv

arxiv.org/abs/2607.12605v1

Read the source

Source

Plain-Language Brief

A new automated program repair framework called CT-Repair is presented, which addresses two limitations of existing large language models in this area. The first issue is the large and repetitive nature of raw execution traces, and the second is that repeated patch sampling may not yield distinct root-cause hypotheses or repair strategies. CT-Repair uses Code Property Graphs (CPG) and Temporal Execution Graphs (TEG) to represent static and dynamic evidence in a queryable format. The framework applies a three-stage filtering pipeline to construct compact TEGs, and three agents analyze each bug from different perspectives to produce evidence-grounded repair strategies. CT-Repair was evaluated on 854 Java bugs from Defects4J v3.0, and it correctly repaired more bugs than ReinFix and RepairAgent in a controlled GPT-5.4-mini configuration.

What the Document Says

  • CT-Repair is an automated program repair framework
  • The framework addresses two limitations of large language models in automated program repair: the large and repetitive nature of raw execution traces, and repeated patch sampling not yielding distinct root-cause hypotheses or repair strategies
  • CT-Repair represents static and dynamic evidence as Code Property Graphs (CPG) and Temporal Execution Graphs (TEG)
  • Three finite-state-machine-guided agents analyze each bug from static, dynamic, and hybrid perspectives
  • A strategy-guided generation procedure instantiates repair strategies as candidate patches and uses validation feedback to refine the most promising strategy
  • In an evaluation on 854 Java bugs from Defects4J v3.0, CT-Repair correctly repaired more bugs than ReinFix and RepairAgent in a controlled GPT-5.4-mini configuration
  • The filtering pipeline compacts runtime evidence by narrowing the candidate method scope by 94.85% on average and reducing retained runtime records by 55.97%

Key Concepts Mentioned

  • automated program repair
  • large language models
  • Code Property Graphs (CPG)
  • Temporal Execution Graphs (TEG)
  • finite-state-machine-guided agents
  • strategy-guided generation
  • Defects4J v3.0

Claims Made by the Document

  • CT-Repair addresses two limitations of large language models in automated program repair
  • In an evaluation on 854 Java bugs from Defects4J v3.0, CT-Repair correctly repaired more bugs than ReinFix and RepairAgent in a controlled GPT-5.4-mini configuration

Evidence Used

  • CT-Repair is presented in the paper
  • The evaluation results are provided in the paper

Methods / Approach

  • Three-stage filtering pipeline
  • Finite-state-machine-guided agents
  • Strategy-guided generation

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

Large language models (LLMs) have improved automated program repair (APR), but two limitations remain. First, raw execution traces are often too large and repetitive to serve as effective model context. Second, repeated patch sampling may produce different implementations without yielding distinct root-cause hypotheses or repair strategies. We present CT-Repair, an agentic APR framework representing static and dynamic evidence as queryable Code Property Graph (CPG) and Temporal Execution Graph (TEG). CT-Repair applies a three-stage filtering pipeline to construct compact TEGs. Three finite-state-machine-guided agents analyze each bug from static, dynamic, and hybrid perspectives and independently produce evidence-grounded repair strategies. A strategy-guided generation procedure instantiates these strategies as candidate patches and uses validation feedback to refine the most promising strategy. We evaluate CT-Repair on 854 Java bugs from Defects4J v3.0. In the mixed-model configuration, CT-Repair correctly repairs 489 bugs. Under a controlled GPT-5.4-mini configuration, it repairs 388 bugs, 19 and 30 more than ReinFix and RepairAgent, respectively. The union of the three evidence perspectives repairs 99 more bugs than the strongest individual perspective. The filtering pipeline also compacts runtime evidence, with execution filtering narrowing the candidate method scope by 94.85% on average and behavior filtering further reducing retained runtime records by 55.97%. These results show that structured runtime evidence and multi-perspective reasoning can improve repair effectiveness without relying solely on a larger patch-generation budget.