We introduce SkillGate, a training method for in-policy skill selection in long-horizon agents. It addresses selector credit starvation by separating outcome credit for execution tokens from an action-local advantage applied only to skill-naming tokens. On five agentic benchmarks with a 16-candidate slate, SkillGate improves a 9B policy from 40.8% to 53.2% trial success while reducing exposure to misleading skills by two thirds.</p>\n","updatedAt":"2026-08-20T09:00:19.998Z","author":{"_id":"64c4ab0388373ea6200e1cf3","avatarUrl":"/avatars/8ad27c35d3def048bc4ff96c0510bba6.svg","fullname":"qingyao li","name":"simonlqy","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9107390642166138},"editors":["simonlqy"],"editorAvatarUrls":["/avatars/8ad27c35d3def048bc4ff96c0510bba6.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.18852","authors":[{"_id":"6a8666b9db13816030683ea9","user":{"_id":"64c4ab0388373ea6200e1cf3","avatarUrl":"/avatars/8ad27c35d3def048bc4ff96c0510bba6.svg","isPro":false,"fullname":"qingyao li","user":"simonlqy","type":"user","name":"simonlqy"},"name":"Qingyao Li","status":"claimed_verified","statusLastChangedAt":"2026-08-20T08:45:04.834Z","hidden":false},{"_id":"6a8666b9db13816030683eaa","name":"Wenxiang Jiao","hidden":false},{"_id":"6a8666b9db13816030683eab","name":"Shuai Shao","hidden":false},{"_id":"6a8666b9db13816030683eac","name":"Kangning Zhang","hidden":false},{"_id":"6a8666b9db13816030683ead","name":"Yuan Lu","hidden":false},{"_id":"6a8666b9db13816030683eae","name":"Yi Guo","hidden":false},{"_id":"6a8666b9db13816030683eaf","name":"Weiwen Liu","hidden":false},{"_id":"6a8666b9db13816030683eb0","name":"Weinan Zhang","hidden":false},{"_id":"6a8666b9db13816030683eb1","name":"Yong Yu","hidden":false}],"publishedAt":"2026-08-19T00:00:00.000Z","submittedOnDailyAt":"2026-08-20T00:00:00.000Z","title":"SkillGate: Training In-Policy Skill Selection in Long-Horizon Agents","submittedOnDailyBy":{"_id":"64c4ab0388373ea6200e1cf3","avatarUrl":"/avatars/8ad27c35d3def048bc4ff96c0510bba6.svg","isPro":false,"fullname":"qingyao li","user":"simonlqy","type":"user","name":"simonlqy"},"summary":"Agent frameworks increasingly package procedural knowledge as skills: instruction files an agent reads on demand, while public libraries now hold thousands of them. Which skill to read has thus become a decision the policy itself makes in the middle of an episode, yet no existing signal trains it. We show that the default remedy, outcome-rewarded RL over the candidate slate, cannot teach it, for a structural reason we identify and name selector credit starvation: under a broadcast, sequence-level advantage, the few tokens that name the chosen skill carry a vanishing share of the loss, and the credit they inherit is increasingly wrong-signed as trajectories lengthen. A correct choice is punished whenever the execution after it fails, even though the choice itself is among the most valuable decisions in the trajectory. Auditing a completed run's own training artifacts confirms all three properties, each worsening monotonically with horizon. SkillGate removes the failure by construction: it partitions the token support into two disjoint credit channels, outcome credit reaching only execution tokens, and a separate action-local advantage reaching exactly the skill-naming tokens, positive only when a trajectory's single read is the correct one. On five agentic benchmarks under a 16-candidate slate, SkillGate lifts a 9B policy from 40.8% to 53.2% trial success, well ahead of the identical budget spent on outcome reward alone, while cutting exposure to misleading candidates by two thirds and reading fewer skills.","upvotes":1,"discussionId":"6a8666b9db13816030683eb2","githubRepo":"https://github.com/DeepExperience/SkillGate","githubRepoAddedBy":"user","ai_summary":"SkillGate fixes selector credit starvation in agent skill selection by separating outcome credit for execution tokens from local advantage for skill-naming tokens, improving success rates and reducing misleading skill exposure.","ai_keywords":["selector credit starvation","outcome-rewarded RL","sequence-level advantage","skill selection","SkillGate","action-local advantage","agentic benchmarks"],"ai_summary_model":"thinkingmachines/Inkling-Small","githubStars":0,"organization":{"_id":"63e5ef7bf2e9a8f22c515654","name":"SJTU","fullname":"Shanghai Jiao Tong University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1676013394657-63e5ee22b6a40bf941da0928.png"}},"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"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"63e5ef7bf2e9a8f22c515654","name":"SJTU","fullname":"Shanghai Jiao Tong University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1676013394657-63e5ee22b6a40bf941da0928.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.18852.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2608.18852 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2608.18852 in a Space README.md to link it from this page.

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