Introducing AURA: memory for robots that knows when to shut up.</p>\n<p>Robots don’t need to remember every frame. They need to remember what changes the next action.</p>\n<p>AURA keeps VLA memory constant at 4,224 bytes and cuts writes up to 9x without hurting closed-loop success.<br><a href=\"https://cdn-uploads.huggingface.co/production/uploads/64442f46af034cdfd69d5bc4/UIDGI6qUjFUbVSB118GXg.jpeg\" rel=\"nofollow\"><img src=\"https://cdn-uploads.huggingface.co/production/uploads/64442f46af034cdfd69d5bc4/UIDGI6qUjFUbVSB118GXg.jpeg\" alt=\"Home Image\"></a></p>\n","updatedAt":"2026-06-03T14:13:46.910Z","author":{"_id":"64442f46af034cdfd69d5bc4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/S1CIatxL_H5QWDA6fgOop.png","fullname":"Josef Chen","name":"josefchen","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8195353746414185},"editors":["josefchen"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/S1CIatxL_H5QWDA6fgOop.png"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.02775","authors":[{"_id":"6a200b40e292c1c78ecb15e9","user":{"_id":"64442f46af034cdfd69d5bc4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/S1CIatxL_H5QWDA6fgOop.png","isPro":true,"fullname":"Josef Chen","user":"josefchen","type":"user","name":"josefchen"},"name":"Josef Chen","status":"claimed_verified","statusLastChangedAt":"2026-06-03T14:17:04.925Z","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/64442f46af034cdfd69d5bc4/0-F_Djke_mY3o68tCNb15.jpeg"],"publishedAt":"2026-06-01T18:38:21.000Z","submittedOnDailyAt":"2026-06-03T00:00:00.000Z","title":"AURA: Action-Gated Memory for Robot Policies at Constant VRAM","submittedOnDailyBy":{"_id":"64442f46af034cdfd69d5bc4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/S1CIatxL_H5QWDA6fgOop.png","isPro":true,"fullname":"Josef Chen","user":"josefchen","type":"user","name":"josefchen"},"summary":"The KV-cache is the right memory for datacenters but the wrong memory for robots. Datacenter inference batches many short requests and resets them, amortizing an attention cache across a crowd. Embodied agents instead run one long, non-resetting episode on bandwidth-limited edge hardware, where high-bandwidth memory and flash are scarce, flash has finite write endurance, and memory writes rather than compute can become the binding constraint.\n AURA-Mem (Action-Utility Recurrent Adaptive Memory) targets this regime. It wraps a frozen vision-language-action backbone with a constant-size recurrent memory and a learned gate that writes only when the current observation would change the next action: memory that knows when to stay silent. Unlike reconstruction-based memory, the gate is trained directly against a closed-loop action-error signal. Its inference state is fixed at 4,224 bytes regardless of horizon, while a KV-cache grows to 6,061 times larger at 100,000 steps.\n On a controlled synthetic benchmark, AURA-Mem matches the best O(1) baseline in accuracy while using 5.19-6.13 times fewer writes, and up to 9.19 times fewer writes on easier configurations. Budget-matched random and periodic schedules do not recover this gain, isolating the benefit to the action-surprise signal. On a trained closed-loop OpenVLA-OFT 7B panel on LIBERO-Long (n=60 episodes per arm), the gate does not hurt success: AURA-Mem matches the ungated base policy (0.233) and slightly exceeds an always-write KV arm (0.217), while using 7.0 times fewer writes and constant memory. We also instantiate an approximate-information-state value-loss bound as a methodology demonstration; at this scale, the bound is vacuous rather than a guarantee.","upvotes":2,"discussionId":"6a200b41e292c1c78ecb15ea","projectPage":"https://huggingface.co/spaces/KAIKAKU/aura-demo","ai_summary":"AURA-Mem is a recurrent memory system that adapts to embodied AI constraints by writing only when observations affect actions, significantly reducing memory writes compared to traditional KV-cache approaches.","ai_keywords":["KV-cache","vision-language-action backbone","recurrent memory","learned gate","action-error signal","closed-loop action-error signal","AURA-Mem","memory writes","embodied agents","attention cache","flash memory","write endurance","action-surprise signal","OpenVLA-OFT","LIBERO-Long"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"64442f5f79e7797ce71c3942","name":"Kaikaku","fullname":"Kaikaku","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/64442f46af034cdfd69d5bc4/SRJTr1BaXthGq1fKLBPm3.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64442f46af034cdfd69d5bc4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/S1CIatxL_H5QWDA6fgOop.png","isPro":true,"fullname":"Josef Chen","user":"josefchen","type":"user"},{"_id":"69bcde629267623e872d9681","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/uT07rNDNnY2Qaazbrkud-.png","isPro":false,"fullname":"木村七海","user":"julianhill45","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"64442f5f79e7797ce71c3942","name":"Kaikaku","fullname":"Kaikaku","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/64442f46af034cdfd69d5bc4/SRJTr1BaXthGq1fKLBPm3.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.02775.md"}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.02775 in a dataset README.md to link it from this page.

Add this paper to a collection to link it from this page.