mlsanityData to 7 Oct 2026

Feedback-Normalized Developer Memory for Reinforcement-Learning Coding Agents: A Safety-Gated MCP Architecture

arXiv 2605.01567 ↗First posted 2 May 2026cs.SE cs.CL cs.LGcited by 0

On the map: LLM agents and tool use → Agent memory systems

Abstract (arXiv)

Large language model (LLM) coding agents increasingly operate over repositories, terminals, tests, and execution traces across long software-engineering episodes. Persistent memory is useful, but static vector stores or generic retrieval-augmented generation (RAG) are insufficient for reinforcement-learning (RL) code development, where small details can alter Bellman targets, terminal masks, gradient flow, or validation claims. This paper presents RL Developer Memory, a local-first, Model Context Protocol (MCP)-native developer-memory architecture for RL coding agents. It treats memory selection as a logged contextual decision process: issue_match ranks candidates and records telemetry, issue_feedback maps raw labels to bounded rewards, and issue_record_resolution links verified resolutions to earlier retrieval events. A deterministic ranker remains deployed, while a contextual-bandit residual policy runs in shadow mode and can affect canary behavior only through conservative off-policy-evaluation (OPE) gates. RL/control memories require theory-to-code metadata and review-gated governance. The system is evaluated on a deterministic 200-case benchmark with RL algorithm bugs, hard negatives, review-gated RL/control cases, and low-risk failures. In the same-commit comparison, deterministic control and full shadow/OPE both achieve 80.0% expected-decision accuracy and 100.0% hard-negative suppression; the full configuration adds learning telemetry rather than accuracy gain. Static validation passed 11/11 checks; dynamic integration passed 10/10 cases. The evidence reports limits: active learned-policy deployment and official-client MCP interoperability are unsupported, live full-configuration latency regresses, and 40 residual non-RL failures remain. The contribution is an auditable memory-control architecture with explicit claim boundaries, not a universal coding-agent improvement claim.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Agent memory systems, one of 16 directions in LLM agents and tool use. Builds on 24 of its 37 references that are in the map. First posted on arXiv 2 May 2026.

Builds on

24 of 37 references.

References · 24

  1. 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
  2. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  3. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  4. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  5. 2023Self-Refine: Iterative Refinement with Self-Feedbackcited by 3,317
  6. 2023SWE-bench: Can Language Models Resolve Real-World GitHub Issues?cited by 3,193
  7. 2020Conservative Q-Learning for Offline Reinforcement Learningcited by 1,777
  8. 2024SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineeringcited by 1,628
  9. 2017Deep Reinforcement Learning that Matterscited by 727
  10. 2024LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memorycited by 656
  11. 2023MemoryBank: Enhancing Large Language Models with Long-Term Memorycited by 539
  12. 2024Training Language Models to Self-Correct via Reinforcement Learningcited by 405
12 more
  1. 2021COMBO: Conservative Offline Model-Based Policy Optimizationcited by 365
  2. 2026MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memorycited by 113
  3. 2023Hyperparameters in Reinforcement Learning and How To Tune Themcited by 58
  4. 2024Normalization and effective learning rates in reinforcement learningcited by 55
  5. 2023On the Effectiveness of Large Language Models in Domain-Specific Code Generationcited by 44
  6. 2023Traces of Memorisation in Large Language Models for Codecited by 26
  7. 2024Latent Reward: LLM-Empowered Credit Assignment in Episodic Reinforcement Learningcited by 22
  8. 2024A Method for Evaluating Hyperparameter Sensitivity in Reinforcement Learningcited by 15
  9. 2024Security Threats in Agentic AI Systemcited by 11
  10. 2024Towards Ethical Personal AI Applications: Practical Considerations for AI Assistants with Long-Term Memorycited by 6
  11. 2019BanditRank: Learning to Rank Using Contextual Banditscited by 3
  12. 2024Stepping Out of the Shadows: Reinforcement Learning in Shadow Modecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Coding Agent Memory Post-training: Unlocking the Memory Potential of Pre-trained File Operations for Long-Horizon Tasks via Reinforcement Learning
  2. 2026Learning When to Remember: Risk-Sensitive Contextual Bandits for Abstention-Aware Memory Retrieval in LLM-Based Coding Agents
  3. 2026CodeScout: An Effective Recipe for Reinforcement Learning of Code Search Agents
  4. 2026LEGO-RL: Harness-Native Reinforcement Learning for Coding Agents
  5. 2026Memory as Infrastructure: Reliability Engineering for Persistent Agent Memory in Months-Long LLM-Assisted Development
  6. 2026PYTHALAB-MERA: Validation-Grounded Memory, Retrieval, and Acceptance Control for Frozen-LLM Coding Agents
  7. 2026CodeMidas: Scaling Agentic Coding RL Environments from Code Itself
  8. 2025Memory-R1: Enhancing Large Language Model Agents to Manage and Utilize Memories via Reinforcement Learning