Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.</p>\n","updatedAt":"2026-05-27T04:32:19.495Z","author":{"_id":"6544b9b646dbdeca34ee5f52","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png","fullname":"Yuyang Hu","name":"namespace-ERI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":4,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8952157497406006},"editors":["namespace-ERI"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.24468","authors":[{"_id":"6a167393e9aa3c8e322db52e","user":{"_id":"6544b9b646dbdeca34ee5f52","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png","isPro":false,"fullname":"Yuyang Hu","user":"namespace-ERI","type":"user","name":"namespace-ERI"},"name":"Yuyang Hu","status":"claimed_verified","statusLastChangedAt":"2026-05-27T07:40:57.095Z","hidden":false},{"_id":"6a167393e9aa3c8e322db52f","name":"Hongjin Qian","hidden":false},{"_id":"6a167393e9aa3c8e322db530","name":"Shuting Wang","hidden":false},{"_id":"6a167393e9aa3c8e322db531","name":"Jiongnan Liu","hidden":false},{"_id":"6a167393e9aa3c8e322db532","name":"Ziliang Zhao","hidden":false},{"_id":"6a167393e9aa3c8e322db533","name":"Jiejun Tan","hidden":false},{"_id":"6a167393e9aa3c8e322db534","name":"Zheng Liu","hidden":false},{"_id":"6a167393e9aa3c8e322db535","name":"Zhicheng Dou","hidden":false}],"publishedAt":"2026-05-23T00:00:00.000Z","submittedOnDailyAt":"2026-05-27T00:00:00.000Z","title":"SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent","submittedOnDailyBy":{"_id":"6544b9b646dbdeca34ee5f52","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png","isPro":false,"fullname":"Yuyang Hu","user":"namespace-ERI","type":"user","name":"namespace-ERI"},"summary":"Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.","upvotes":2,"discussionId":"6a167394e9aa3c8e322db536","githubRepo":"https://github.com/qhjqhj00/cabeza","githubRepoAddedBy":"user","ai_summary":"Long-horizon agentic reasoning is enhanced through a state-adaptive memory framework that dynamically manages interaction histories by creating compact memory cues while preserving detailed trajectories for targeted retrieval.","ai_keywords":["long-horizon agentic reasoning","large language models","interaction history","state-adaptive memory","memory cues","trajectory pages","intent-driven recall","expert-guided supervision","reinforcement learning","trajectory-level utility"],"githubStars":3},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6544b9b646dbdeca34ee5f52","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6544b9b646dbdeca34ee5f52/nRx6m1C4wfZ_xSWoBUNJf.png","isPro":false,"fullname":"Yuyang Hu","user":"namespace-ERI","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.24468.md"}">
SAM: State-Adaptive Memory for Long-Horizon Reasoning Agent
Abstract
Long-horizon agentic reasoning is enhanced through a state-adaptive memory framework that dynamically manages interaction histories by creating compact memory cues while preserving detailed trajectories for targeted retrieval.
AI-generated summary
Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.
Community
Long-horizon agentic reasoning requires large language models to act over long interaction histories containing thoughts, tool calls, observations, and partial conclusions. The challenge is not merely that these histories grow long, but that information needed for the current decision may be scattered across distant steps and only become relevant later. Existing approaches address this difficulty by truncating the interaction history, compressing it into shorter surrogates, or retrieving selected parts of it for reuse, but they do not explicitly model how access to past interaction should adapt to the agent's evolving state. We instead cast long-horizon reasoning as a problem of state-adaptive memory. To this end, we propose State-Adaptive Memory~(SAM), a standalone framework that consolidates ongoing interaction into compact memory cues while preserving raw trajectory pages for intent-driven recall. These cues are not treated as replacements for history; rather, they serve as lightweight handles that allow the agent to reconstruct temporally distant information according to its current needs, without retraining the underlying backbone. We further optimize the memory module through expert-guided supervision and reinforcement learning, aligning it with trajectory-level utility. Across BrowseComp, BrowseComp-ZH, WideSearch, and HLE, SAM consistently outperforms strong baselines over diverse agent backbones. Our results suggest that explicit memory modeling provides a simple and effective foundation for long-horizon agentic reasoning.
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/2605.24468 in a model README.md to link it from this page.
Cite arxiv.org/abs/2605.24468 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2605.24468 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.