[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-new-tool-menu-design-boosts-ai-agent-success-rate-on-toolbench":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},6283,"new-tool-menu-design-boosts-ai-agent-success-rate-on-toolbench","New Tool Menu Design Boosts AI Agent Success Rate on ToolBench","A new tool-selection method for AI agents lifts task success from 74% to 90% on a standard benchmark by ordering prerequisite tools before final actions.","A new paper argues that AI agents don't need bigger tool libraries - they need better ordered ones.\n\nResearchers built a system that hands an agent a short, ranked menu of tools before it starts a task, rather than a giant list ranked purely by relevance to the request. The method, called State-Path Tool Menu, tracks which tools can run from the agent's current state, which tool outputs feed later steps, and which sequences show up repeatedly in training data. On the ToolBench benchmark, that approach raised the agent's online task success rate from 73.7% to 89.8%, without touching the underlying language model. It also matched or beat the coverage of ToolBench's official 128-tool list using a menu of just 32 tools, and the gains held across agents built on different model sizes.\n\nThat's a bigger deal than it sounds. Most agent tooling work chases relevance - surfacing the tool that matches the request - which quietly assumes the final action is the hard part. This paper's real find is that the hard part is often the setup: the boring prerequisite calls that fetch an ID or a file before the agent can do the thing it actually wants to do. Treating tool selection as a routing problem, not a search problem, is a cleaner fix than just cramming more tools into context.\n\nStill, ToolBench is a benchmark, not a production API catalog with rate limits, auth quirks, and tools that silently change behavior. Cleaner numbers on a controlled test don't guarantee the approach survives contact with a real enterprise tool library.","[\"ai agents\",\"tool use\",\"llm research\",\"toolbench\"]","2026-09-11T04:00:00.000Z","2026-09-11T04:21:13.373Z","2026-09-11T04:21:25.292Z","published",null,[],"ai",[26,27,28,29],"ai agents","tool use","llm research","toolbench",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.09395",0,{"sections":36},[37,40,44,48,53,58,63,66,71,75,80,85,90,95],{"name":38,"slug":24,"count":39,"latest_published_at":18},"AI",3507,{"name":41,"slug":42,"count":43,"latest_published_at":18},"Security","security",636,{"name":45,"slug":46,"count":47,"latest_published_at":18},"Policy","policy",338,{"name":49,"slug":50,"count":51,"latest_published_at":52},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":54,"slug":55,"count":56,"latest_published_at":57},"Hardware","hardware",153,"2026-09-09T15:12:32.000Z",{"name":59,"slug":60,"count":61,"latest_published_at":62},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":64,"slug":65,"count":61,"latest_published_at":18},"Science","science",{"name":67,"slug":68,"count":69,"latest_published_at":70},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":72,"slug":73,"count":74,"latest_published_at":18},"Dev Tools","dev-tools",70,{"name":76,"slug":77,"count":78,"latest_published_at":79},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"General","general",41,"2026-09-08T01:57:23.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]