読書リサーチレーダー投資フレームワーク
ログイン / 新規登録
← 読解に戻る
ニューストピック · MemTrace记忆系统重大1 分で読めるDisconfirmAI 編集部

MemTrace Preprint: State-Consistent Memory Boosts DeepSWE Pass@1 by 21.2 Points for Coding Agents

New memory system validates historical evidence against current repo state, significantly outperforming baselines in long-horizon coding tasks.

2026-10-06 12:00 GMT+8

The MemTrace system improves the pass@1 metric on the DeepSWE benchmark by 21.2 percentage points.

Traditional coding agents often fail to reconstruct a consistent task state during long-horizon, multi-file tasks because context refreshes or repository changes invalidate earlier execution evidence.

MemTrace stores execution history as immutable 'Memory Traces' anchored to key information and checks the validity of historical evidence against the current repository state before restoring it. Under the Codex CLI framework, compared to all fully evaluated baselines, it improved the SWE-EVO Resolved Rate by 4.4 points and the SWE-Milestone Score by 17.8 points.

Data comes from author-run tests in a preprint (v2 updated Oct 8); independent reproduction has not yet been reported.

ソース

  • arxiv.org↗
トピック

MemTrace记忆系统

【背景】近期两项名为“MemTrace”的研究分别解决了AI记忆系统的两大痛点:一是长程编码中的状态一致性,二是记忆错误的可追溯性。

3 ステップ2 件更新日 2026-10-10 13:21

経緯

ログイン / 新規登録
読書リサーチレーダー投資フレームワーク
2026-05-28
浙大与阿里团队发布《MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems》,提出LLM记忆错误归因框架及MemTraceBench基准。
arxiv.org ↗
  • 2026-10-04Hongming Xu等人提交预印本《MemTrace: State-Consistent Memory for Long-Horizon Coding Agents》,介绍面向编码智能体的状态一致记忆系统。arxiv.org ↗
  • 2026-10-08编码版MemTrace论文更新v2版本,确认在DeepSWE基准上提升21.2点,并在SWE-EVO等指标上优于基线。arxiv.org ↗
  • 未解決の問題

    • 编码版MemTrace的“记忆轨迹图”在超大规模代码库中的索引开销与实时性表现如何?
    • 错误归因版MemTrace目前最佳的操作识别准确率不足50%,如何进一步提升对复杂并发错误的定位精度?
    • 两种MemTrace方法是否可以在统一的智能体架构中结合,既保证状态一致性又提供可解释的错误调试能力?

    このスレッド内の他の記事

    • MemTrace Boosts Long-Horizon Coding Agents2026-10-07
    トピック全文を読む →
    リサーチを続ける →