Mem-π is an adaptive memory framework for LLM agents that replaces traditional retrieval with a model-generated, RL-optimized guidance mechanism to improve performance on complex, context-dependent agentic tasks.</p>\n","updatedAt":"2026-05-21T02:26:35.767Z","author":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","fullname":"taesiri","name":"taesiri","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":302,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8433855772018433},"editors":["taesiri"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.21463","authors":[{"_id":"6a0e6d54164dbbc68a26c473","name":"Xiaoqiang Wang","hidden":false},{"_id":"6a0e6d54164dbbc68a26c474","name":"Chao Wang","hidden":false},{"_id":"6a0e6d54164dbbc68a26c475","name":"Hadi Nekoei","hidden":false},{"_id":"6a0e6d54164dbbc68a26c476","name":"Christopher Pal","hidden":false},{"_id":"6a0e6d54164dbbc68a26c477","name":"Alexandre Lacoste","hidden":false},{"_id":"6a0e6d54164dbbc68a26c478","name":"Spandana Gella","hidden":false},{"_id":"6a0e6d54164dbbc68a26c479","name":"Bang Liu","hidden":false},{"_id":"6a0e6d54164dbbc68a26c47a","name":"Perouz Taslakian","hidden":false}],"publishedAt":"2026-05-20T00:00:00.000Z","submittedOnDailyAt":"2026-05-21T00:00:00.000Z","title":"Mem-π: Adaptive Memory through Learning When and What to Generate","submittedOnDailyBy":{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user","name":"taesiri"},"summary":"We present Mem-π, a framework for adaptive memory in large language model (LLM) agents, where useful guidance is generated on demand rather than retrieved from external memory stores. Existing memory-augmented agents typically rely on similarity-based retrieval from episodic memory banks or skill libraries, returning static entries that often misalign with the current context. In contrast, Mem-π uses a dedicated language or vision-language model with its own parameters, separate from the downstream agent, to generate context-specific guidance for complex tasks. Conditioned on the current agent context, the model jointly decides when to produce guidance and what guidance to produce. We train it with a decision-content decoupled reinforcement learning (RL) objective, enabling it to abstain when generation would not help and otherwise produce concise, useful guidance. Across diverse agentic benchmarks spanning web navigation, terminal-based tool use, and text-based embodied interaction, Mem-π consistently outperforms retrieval-based and prior RL-optimized memory baselines, achieving over 30% relative improvement on web navigation tasks.","upvotes":2,"discussionId":"6a0e6d55164dbbc68a26c47b","ai_summary":"Mem-π is a framework for adaptive memory in LLM agents that generates context-specific guidance using a separate language or vision-language model trained with decision-content decoupled reinforcement learning.","ai_keywords":["large language model agents","adaptive memory","episodic memory banks","skill libraries","similarity-based retrieval","language model","vision-language model","reinforcement learning","decision-content decoupled reinforcement learning","web navigation","terminal-based tool use","text-based embodied interaction"]},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"699edcd4796be456dc1dfaf7","avatarUrl":"/avatars/7d9112dbd6e723b7c945f542ac619782.svg","isPro":false,"fullname":"杨紫瑜","user":"ELIJAHHI3","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.21463.md"}">
Mem-π: Adaptive Memory through Learning When and What to Generate
Abstract
Mem-π is a framework for adaptive memory in LLM agents that generates context-specific guidance using a separate language or vision-language model trained with decision-content decoupled reinforcement learning.
AI-generated summary
We present Mem-π, a framework for adaptive memory in large language model (LLM) agents, where useful guidance is generated on demand rather than retrieved from external memory stores. Existing memory-augmented agents typically rely on similarity-based retrieval from episodic memory banks or skill libraries, returning static entries that often misalign with the current context. In contrast, Mem-π uses a dedicated language or vision-language model with its own parameters, separate from the downstream agent, to generate context-specific guidance for complex tasks. Conditioned on the current agent context, the model jointly decides when to produce guidance and what guidance to produce. We train it with a decision-content decoupled reinforcement learning (RL) objective, enabling it to abstain when generation would not help and otherwise produce concise, useful guidance. Across diverse agentic benchmarks spanning web navigation, terminal-based tool use, and text-based embodied interaction, Mem-π consistently outperforms retrieval-based and prior RL-optimized memory baselines, achieving over 30% relative improvement on web navigation tasks.
Community
Mem-π is an adaptive memory framework for LLM agents that replaces traditional retrieval with a model-generated, RL-optimized guidance mechanism to improve performance on complex, context-dependent agentic tasks.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2605.21463 in a model README.md to link it from this page.
Cite arxiv.org/abs/2605.21463 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2605.21463 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.