# Learning to Commit: Generating Organic Pull Requests via Online Repository Memory

arXiv: https://arxiv.org/abs/2603.26664 · HTML: https://mlsanity.com/paper/2603.26664/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use.
- Builds on 19 of its 21 references that are in the map.
- First posted on arXiv 27 Mar 2026.

## Placed in

- LLM agents and tool use / [Software engineering coding agents](https://mlsanity.com/agents/software-engineering-coding-agents/)

## Builds on (most cited first, 19 of 19)

- [Evaluating Large Language Models Trained on Code](https://mlsanity.com/paper/2107.03374/) (2021)
- [Program Synthesis with Large Language Models](https://mlsanity.com/paper/2108.07732/) (2021)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [R2E-Gym: Procedural Environments and Hybrid Verifiers for Scaling Open-Weights SWE Agents](https://mlsanity.com/paper/2504.07164/) (2025)
- [OpenClaw-RL: Train Any Agent Simply by Talking](https://mlsanity.com/paper/2603.10165/) (2026)
- [Memory for Autonomous LLM Agents:Mechanisms, Evaluation, and Emerging Frontiers](https://mlsanity.com/paper/2603.07670/) (2026)
- [FEA-Bench: A Benchmark for Evaluating Repository-Level Code Generation for Feature Implementation](https://mlsanity.com/paper/2503.06680/) (2025)
- [SWE-Bench++: A Framework for the Scalable Generation of Software Engineering Benchmarks from Open-Source Repositories](https://mlsanity.com/paper/2512.17419/) (2025)
- [SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via Continuous Integration](https://mlsanity.com/paper/2603.03823/) (2026)
- [Immersion in the GitHub Universe: Scaling Coding Agents to Mastery](https://mlsanity.com/paper/2602.09892/) (2026)
- [Improving Code Localization with Repository Memory](https://mlsanity.com/paper/2510.01003/) (2025)
- [Structurally Aligned Subtask-Level Memory for Software Engineering Agents](https://mlsanity.com/paper/2602.21611/) (2026)
- [Coeditor: Leveraging Contextual Changes for Multi-round Code Auto-editing](https://mlsanity.com/paper/2305.18584/) (2023)
- [SWE Context Bench: A Benchmark for Context Learning in Coding](https://mlsanity.com/paper/2602.08316/) (2026)
- [SGAgent: Suggestion-Guided LLM-Based Multi-Agent Framework for Repository-Level Software Repair](https://mlsanity.com/paper/2602.23647/) (2026)
- [Live-Evo: Online Evolution of Agentic Memory from Continuous Feedback](https://mlsanity.com/paper/2602.02369/) (2026)
- [Agyn: A Multi-Agent System for Team-Based Autonomous Software Engineering](https://mlsanity.com/paper/2602.01465/) (2026)
- [CommitBART: A Large Pre-trained Model for GitHub Commits](https://mlsanity.com/paper/2208.08100/) (2022)
- [Pull Requests as a Training Signal for Repo-Level Code Editing](https://mlsanity.com/paper/2602.07457/) (2026)

## Built on by (most cited first, 3 of 3)

- [Code as Agent Harness](https://mlsanity.com/paper/2605.18747/) (2026)
- [ContextSniper: AntTrail's Token-Efficient Code Memory for Repository-Level Program Repair](https://mlsanity.com/paper/2607.01916/) (2026)
- [Skill Training with Corruption and Reconstruction Loop](https://mlsanity.com/paper/2607.27557/) (2026)

## Abstract (arXiv)

Large language model (LLM)-based coding agents achieve impressive results on controlled benchmarks yet routinely produce pull requests that real maintainers reject. The root cause is not functional incorrectness but a lack of organicity: generated code ignores project-specific conventions, duplicates functionality already provided by internal APIs, and violates implicit architectural constraints accumulated over years of development. Simply exposing an agent to the latest repository snapshot is not enough: the snapshot reveals the final state of the codebase, but not the repository-specific change patterns by which that state was reached. We introduce Learning to Commit, a framework that closes this gap through Online Repository Memory. Given a repository with a strict chronological split, the agent performs supervised contrastive reflection on earlier commits: it blindly attempts to resolve each historical issue, compares its prediction against the oracle diff, and distils the gap into a continuously growing set of skills-reusable patterns capturing coding style, internal API usage, and architectural invariants. When a new PR description arrives, the agent conditions its generation on these accumulated skills, producing changes grounded in the project's own evolution rather than generic pretraining priors. Evaluation is conducted on genuinely future, merged pull requests that could not have been seen during the skill-building phase, and spans multiple dimensions including functional correctness, code-style consistency, internal API reuse rate, and modified-region plausibility. Experiments on an expert-maintained repository with rich commit history show that Online Repository Memory effectively improves organicity scores on held-out future tasks.