Chronos Enables Code Agents to Reason over Software Evolution

2026-10-08

AI Summary

This study addresses the challenge that coding agents struggle to reuse design decisions and compatibility constraints from historical pull requests (PRs), resulting in a lack of experiential knowledge. To overcome this, the authors propose a test-time framework that distills merged PRs into structured experience cards and constructs a typed relational graph integrating code, intent, and organizational dependencies. Furthermore, the method introduces graph-based semantic retrieval with multi-hop expansion mechanisms, alongside an evolutionary steward module to guide patch generation and selection. Evaluated on SWE-Bench Verified, the approach achieves an average resolution rate of 72.9% (peaking at 79.8%) and increases effective card retrieval by over 130%, demonstrating precise reuse of historical development experience.

Abstract

Historical pull requests record the design decisions, compatibility constraints, and implementation patterns behind a codebase's current state. Experience relevant to a new task can span related changes whose descriptions emphasize different concerns. We introduce Chronos, a test-time framework that makes this connected history available to large language model (LLM)-based code agents. Chronos distills merged pull requests into structured experience cards and connects them through a typed graph of code-level, developer-intent, and organizational relations. Semantic search identifies entry cards, and weighted multi-hop expansion retrieves connected changes for selective reading. The same memory guides candidate generation and patch selection: a patch-focused change agent and a validation-strategy agent each develop a patch, and an evolution steward consults history to select between them. On SWE-Bench Verified, the full workflow improves SWE-Agent across all six evaluated LLM backbones, raising the mean resolution rate from 69.2% to 72.9% and reaching 79.8% with MiniMax M2.5. With the same backbone, it raises resolution rates from 48.3% to 51.7% on SWE-Bench Pro and from 41.0% to 43.5% on FEA-Bench Lite. Both experience-guided single-agent variants also outperform the base agent. In a human evaluation on 100 tasks with ten cards retrieved per task, graph-grounded retrieval increases the mean number of useful cards from 1.24 to 2.87 over flat semantic retrieval. These results demonstrate the value of PR relations for retrieving useful repository experience and of the evaluated workflows for applying that experience during patch generation and selection.

Problem

• Code Agents
• Software Evolution
• Pull Requests
• Large Language Models
• Repository Experience

Innovation

• Code Agents
• Pull Request Graph
• Experience Cards
• Multi-hop Retrieval
• Multi-Agent Workflow