Agora: Git as Shared Memory for Collective AutoResearch</p>\n","updatedAt":"2026-09-17T04:00:33.135Z","author":{"_id":"647bf082aba7062fe5c51ca9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/647bf082aba7062fe5c51ca9/VvKAhQC_LxBcBuy3XROSX.jpeg","fullname":"Yifan Zhang","name":"yifAI","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":32,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8690199851989746},"editors":["yifAI"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/647bf082aba7062fe5c51ca9/VvKAhQC_LxBcBuy3XROSX.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2609.18094","authors":[{"_id":"6aab65c11d9cc4dec79624c2","user":{"_id":"647bf082aba7062fe5c51ca9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/647bf082aba7062fe5c51ca9/VvKAhQC_LxBcBuy3XROSX.jpeg","isPro":true,"fullname":"Yifan Zhang","user":"yifAI","type":"user","name":"yifAI"},"name":"Yifan Zhang","status":"claimed_verified","statusLastChangedAt":"2026-09-17T08:45:04.504Z","hidden":false},{"_id":"6aab65c11d9cc4dec79624c3","name":"Yunheng Zou","hidden":false},{"_id":"6aab65c11d9cc4dec79624c4","name":"Shaokun Zhang","hidden":false},{"_id":"6aab65c11d9cc4dec79624c5","name":"Jian Hu","hidden":false},{"_id":"6aab65c11d9cc4dec79624c6","name":"Hao Zhang","hidden":false},{"_id":"6aab65c11d9cc4dec79624c7","name":"Binfeng Xu","hidden":false},{"_id":"6aab65c11d9cc4dec79624c8","name":"Jan Kautz","hidden":false},{"_id":"6aab65c11d9cc4dec79624c9","name":"Yi Dong","hidden":false}],"publishedAt":"2026-09-16T00:00:00.000Z","submittedOnDailyAt":"2026-09-17T00:00:00.000Z","title":"Agora: Git as Shared Memory for Collective AutoResearch","submittedOnDailyBy":{"_id":"647bf082aba7062fe5c51ca9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/647bf082aba7062fe5c51ca9/VvKAhQC_LxBcBuy3XROSX.jpeg","isPro":true,"fullname":"Yifan Zhang","user":"yifAI","type":"user","name":"yifAI"},"summary":"Autonomous research loops such as AutoResearch show that one coding agent can improve a training setup unattended. Run several of them and each session starts from scratch, so more agents tend to mean more duplicated search rather than more discovery. Agora is a shared memory for such agents: research is recorded as an append-only directed acyclic graph (DAG) stored in Git, so that every claim is a commit anyone can check out and rerun. Each result, insight, hypothesis, verification, and report is an immutable commit whose parent edges say what it builds on; a derived index exposes the frontier, the neglected branches, and the verification status of each claim, and a diversity-aware selection rule keeps the community from collapsing onto one leader. We describe the system and report its first sustained use: a run of nearly 12 days in which 13 language-model workers, with no assigned tasks and no central planner, worked on a weight-transfer problem. Given 141 pretrained donor models and a frozen 119.6M-parameter attention-SSM hybrid whose dimensions match no donor, the workers had to initialize the target without training data or gradient updates. They published 1,703 contributions and drove the evaluator from 3.39 to 1.899 bits per byte, closing 62% of the gap to a trained GPT-2 124M. The winning recipe compresses donor next-token statistics into the target's embedding and output head, then adds a short-range context signal through sparse edits to attention, feed-forward, and state-space blocks. Its 145-commit ancestry spans 15 accounts, and 165 independent reproductions were posted, none of which failed. We describe the single mid-run human intervention that pulled the community out of a monoculture, what the trace does and does not establish, and the controlled comparison that would settle whether shared research state improves discovery per unit of compute.","upvotes":35,"discussionId":"6aab65c21d9cc4dec79624ca","organization":{"_id":"60262b67268c201cdc8b7d43","name":"nvidia","fullname":"NVIDIA","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65df9200dc3292a8983e5017/Vs5FPVCH-VZBipV3qKTuy.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"647bf082aba7062fe5c51ca9","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/647bf082aba7062fe5c51ca9/VvKAhQC_LxBcBuy3XROSX.jpeg","isPro":true,"fullname":"Yifan Zhang","user":"yifAI","type":"user"},{"_id":"65574b4e60cb377db0546c15","avatarUrl":"/avatars/7fb7f10cd1eda32270c96a0cc871c2be.svg","isPro":false,"fullname":"Jeff Peng","user":"jeffpeng","type":"user"},{"_id":"691fed991ab51cfbbbf8386e","avatarUrl":"/avatars/777e5bd3ae4c76e703bdbc35abfd7192.svg","isPro":false,"fullname":"Jian Hu","user":"jianh-nvidia","type":"user"},{"_id":"645b0cb3333fb18357875c96","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/645b0cb3333fb18357875c96/X-JiamCXh0bbcAfVvRpkj.jpeg","isPro":true,"fullname":"Binfeng Xu","user":"billxbf","type":"user"},{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user"},{"_id":"66d8512c54209e9101811e8e","avatarUrl":"/avatars/62dfd8e6261108f2508efe678d5a2a57.svg","isPro":false,"fullname":"M Saad Salman","user":"MSS444","type":"user"},{"_id":"69a399802c41112bc82e8a84","avatarUrl":"/avatars/074014e6da2ebe60231731b9006b181f.svg","isPro":false,"fullname":"LUO MINCHEN","user":"mandyluo","type":"user"},{"_id":"670fee526b8de4974727e86d","avatarUrl":"/avatars/cb802613069a1830b5f3f9c05f5a0603.svg","isPro":false,"fullname":"Zhuoqing Song","user":"Mat3579","type":"user"},{"_id":"6801df6171a387d10b3ba93c","avatarUrl":"/avatars/0cbfd1ec6125b4e0441db48e1b9677a6.svg","isPro":false,"fullname":"Shaokun zhang","user":"SeanZhang1","type":"user"},{"_id":"6a6a92931b3822fa45f64623","avatarUrl":"/avatars/c51da7089b42b77a78ee76de22250514.svg","isPro":false,"fullname":"Jessica Brown","user":"jessica-brown","type":"user"},{"_id":"6a6a9d781fcc333c083e0a2f","avatarUrl":"/avatars/31b1c26caa8af0dd423cc1ff6ed9af43.svg","isPro":false,"fullname":"Brian Jones","user":"brian-jones","type":"user"},{"_id":"6a6c8070ef16c968823e47e3","avatarUrl":"/avatars/fce97db08e578c90c99f8c06453ae1bb.svg","isPro":false,"fullname":"Matthew Moore","user":"orbitReed","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"60262b67268c201cdc8b7d43","name":"nvidia","fullname":"NVIDIA","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/65df9200dc3292a8983e5017/Vs5FPVCH-VZBipV3qKTuy.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2609/2609.18094.md","query":{}}">
Agora: Git as Shared Memory for Collective AutoResearch
Abstract
Autonomous research loops such as AutoResearch show that one coding agent can improve a training setup unattended. Run several of them and each session starts from scratch, so more agents tend to mean more duplicated search rather than more discovery. Agora is a shared memory for such agents: research is recorded as an append-only directed acyclic graph (DAG) stored in Git, so that every claim is a commit anyone can check out and rerun. Each result, insight, hypothesis, verification, and report is an immutable commit whose parent edges say what it builds on; a derived index exposes the frontier, the neglected branches, and the verification status of each claim, and a diversity-aware selection rule keeps the community from collapsing onto one leader. We describe the system and report its first sustained use: a run of nearly 12 days in which 13 language-model workers, with no assigned tasks and no central planner, worked on a weight-transfer problem. Given 141 pretrained donor models and a frozen 119.6M-parameter attention-SSM hybrid whose dimensions match no donor, the workers had to initialize the target without training data or gradient updates. They published 1,703 contributions and drove the evaluator from 3.39 to 1.899 bits per byte, closing 62% of the gap to a trained GPT-2 124M. The winning recipe compresses donor next-token statistics into the target's embedding and output head, then adds a short-range context signal through sparse edits to attention, feed-forward, and state-space blocks. Its 145-commit ancestry spans 15 accounts, and 165 independent reproductions were posted, none of which failed. We describe the single mid-run human intervention that pulled the community out of a monoculture, what the trace does and does not establish, and the controlled comparison that would settle whether shared research state improves discovery per unit of compute.
Community
Agora: Git as Shared Memory for Collective AutoResearch
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/2609.18094 in a model README.md to link it from this page.
Cite arxiv.org/abs/2609.18094 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2609.18094 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.