We've upgraded our Interactive Training library, which allows agents (and humans) to adjust optimization parameters and gradients on the fly during training. Check out the live demo: <a href=\"https://interactivetraining.ai/live\" rel=\"nofollow\">https://interactivetraining.ai/live</a></p>\n","updatedAt":"2026-07-27T17:50:52.640Z","author":{"_id":"63081e15a670ed10f9d44229","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63081e15a670ed10f9d44229/w1b9uq-9774bMMgJbSPsS.jpeg","fullname":"Yuntian Deng","name":"yuntian-deng","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":273,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8574252724647522},"editors":["yuntian-deng"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/63081e15a670ed10f9d44229/w1b9uq-9774bMMgJbSPsS.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2607.18314","authors":[{"_id":"6a6798da4534e38b50e9b19c","name":"Wentao Zhang","hidden":false},{"_id":"6a6798da4534e38b50e9b19d","name":"Xuanhe Pan","hidden":false},{"_id":"6a6798da4534e38b50e9b19e","name":"Han Zhou","hidden":false},{"_id":"6a6798da4534e38b50e9b19f","name":"Yang Lu","hidden":false},{"_id":"6a6798da4534e38b50e9b1a0","name":"Yuntian Deng","hidden":false}],"mediaUrls":["https://cdn-uploads.huggingface.co/production/uploads/63081e15a670ed10f9d44229/ql4SDCjCH2GTQvPFZuI-K.png"],"publishedAt":"2026-07-17T00:00:00.000Z","submittedOnDailyAt":"2026-07-27T00:00:00.000Z","title":"Interactive Training 2: Auditable Control Plane for Live Model Training","submittedOnDailyBy":{"_id":"63081e15a670ed10f9d44229","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63081e15a670ed10f9d44229/w1b9uq-9774bMMgJbSPsS.jpeg","isPro":true,"fullname":"Yuntian Deng","user":"yuntian-deng","type":"user","name":"yuntian-deng"},"summary":"Experiment trackers show how training is progressing, but changing a live run still usually requires trainer-specific code. We present Interactive Training 2, an open-source control plane for steering training through a shared protocol. Training applications declare which settings and actions they expose, humans and automated controllers submit requests through the same interface, and the training loop validates and applies them at safe control points. A customized Aim workspace combines live metrics and controls with a chronological record of requests and outcomes. We demonstrate the system across five NLP and reinforcement-learning workflows. The released code and traces provide a reusable foundation for auditable human- and agent-guided training.","upvotes":15,"discussionId":"6a6798da4534e38b50e9b1a1","projectPage":"https://interactivetraining.ai/live","organization":{"_id":"6230c3ced93e84e2338765f3","name":"UWaterloo","fullname":"University of Waterloo","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1647363004040-6230c34bacb94a81eec91ed5.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"63081e15a670ed10f9d44229","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63081e15a670ed10f9d44229/w1b9uq-9774bMMgJbSPsS.jpeg","isPro":true,"fullname":"Yuntian Deng","user":"yuntian-deng","type":"user"},{"_id":"68f1a4e088e256c4ab477152","avatarUrl":"/avatars/e40704b1ebb1b78ccc9577121300793a.svg","isPro":false,"fullname":"Abram Durret","user":"abramdurret","type":"user"},{"_id":"6a481e62a0d7ab44609d9289","avatarUrl":"/avatars/7c4e28f3d231c075e9dd91bd5dfb4ca4.svg","isPro":false,"fullname":"Adji Dieng","user":"adjidieng","type":"user"},{"_id":"6a4c1168351c59d0308a9044","avatarUrl":"/avatars/43ffc20406d46be7846bda1162439304.svg","isPro":false,"fullname":"Alex Bruce","user":"alexbruce2","type":"user"},{"_id":"6a4c1261eb180145e39e8ddf","avatarUrl":"/avatars/4b9e8fbd4b5cde90d4bce7a8a747f5d0.svg","isPro":false,"fullname":"Alice Oh","user":"aliceoh2","type":"user"},{"_id":"6a4a6e061f87a121fd75240c","avatarUrl":"/avatars/a7270b47de35c46ce98a9ea15cb55e5e.svg","isPro":false,"fullname":"Bob Xie","user":"bob22223","type":"user"},{"_id":"6875c5d3602bdfd2d122b6b9","avatarUrl":"/avatars/3ed27ab0936ba3c14f13cbca986950dd.svg","isPro":false,"fullname":"chen chen","user":"chenchen2","type":"user"},{"_id":"6a4a6efc87d0c5e9cae1ed71","avatarUrl":"/avatars/d5551de6b7ff23da273d9d7ad49cd0ba.svg","isPro":false,"fullname":"Chris Dyer","user":"chrisdyer2","type":"user"},{"_id":"6a4b652021cffc4d2e903e13","avatarUrl":"/avatars/9c501e3f12b59d70112c468f00bcc604.svg","isPro":false,"fullname":"Chris Manning","user":"chris2maning","type":"user"},{"_id":"6a481cf4a8d48405de73c128","avatarUrl":"/avatars/706b16438d146736c8775e474d649f49.svg","isPro":false,"fullname":"wildcursor2","user":"wildcursor2","type":"user"},{"_id":"68f1a3bd754424101705a4c4","avatarUrl":"/avatars/4e8b1968bacc01a892f57f2f6e3c08d4.svg","isPro":false,"fullname":"Darius Darius","user":"dariusdarius2","type":"user"},{"_id":"68f1a44f1ffea4469b9de690","avatarUrl":"/avatars/f9fcc27efc47cf2e16bd602cb8d4b5c3.svg","isPro":false,"fullname":"Deborah Goyal","user":"debroahgoyal","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"6230c3ced93e84e2338765f3","name":"UWaterloo","fullname":"University of Waterloo","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1647363004040-6230c34bacb94a81eec91ed5.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2607/2607.18314.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2607.18314 in a model README.md to link it from this page.

Cite arxiv.org/abs/2607.18314 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2607.18314 in a Space README.md to link it from this page.

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