The Year in AI Papers
The year/Independent research

Paper 2601.06789

MemGovern: Enhancing Code Agents through Learning from Governed Human Experiences

Qihao Wang, Ziming Cheng, Shuo Zhang, Fan Liu, Rui Xu, Heng Lian, Kunyi Wang, Xiaoming Yu, Jianghao Yin, Sen Hu, Yue Hu, Shaolei Zhang, et al.

Published
Jan 2026
Research lab
Independent
Citations
8
GitHub
Not linked

01 In brief

Summary

MemGovern is a framework that transforms raw GitHub issue-tracking data into structured, agent-friendly experiential memory to enhance autonomous software engineering (SWE) agents.

It addresses the 'closed-world' limitation of agents that ignore historical human debugging experience.

MemGovern uses experience governance to filter, standardize, and quality-control data into 135K experience cards, each with an Index Layer (for retrieval) and a Resolution Layer (for repair logic).

It also introduces agentic experience search with dual primitives (Searching and Browsing) for progressive, logic-driven retrieval.

On SWE-bench Verified, MemGovern improves resolution rates by an average of 4.65% across seven LLM backbones, with gains up to +9.4% for GPT-4o.

Ablations show that larger memory sizes and governed (vs.

raw) experiences yield consistent improvements, and agentic search outperforms static and adaptive RAG.

The framework is plug-and-play, integrating with SWE-Agent, and incurs modest token overhead.

Case studies demonstrate that MemGovern enables semantically correct fixes (e.g., preserving API contracts) compared to baseline agents' superficial patches.

02 From the paper

Abstract

While autonomous software engineering (SWE) agents are reshaping programming paradigms, they currently suffer from a "closed-world" limitation: they attempt to fix bugs from scratch or solely using local context, ignoring the immense historical human experience available on platforms like GitHub. Accessing this open-world experience is hindered by the unstructured and fragmented nature of real-world issue-tracking data. In this paper, we introduce MemGovern, a framework designed to govern and transform raw GitHub data into actionable experiential memory for agents. MemGovern employs experience governance to convert human experience into agent-friendly experience cards and introduces an agentic experience search strategy that enables logic-driven retrieval of human expertise. By producing 135K governed experience cards, MemGovern achieves a significant performance boost, improving resolution rates on the SWE-bench Verified by 4.65%. As a plug-in approach, MemGovern provides a solution for agent-friendly memory infrastructure.