{"paper":{"title":"ClawGym: A Scalable Framework for Building Effective Claw Agents","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"ClawGym provides a complete framework for synthesizing data, training, and evaluating Claw-style personal agents.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bryan Dai, Chuan Hao, Daixuan Cheng, Fei Bai, Feng Chang, Huatong Song, Jian Yang, Ji-Rong Wen, Ran Tao, Renyuan Li, Shuang Sun, Wayne Xin Zhao, Yike Yang, Yuan Wei","submitted_at":"2026-04-29T17:12:22Z","abstract_excerpt":"Claw-style environments support multi-step workflows over local files, tools, and persistent workspace states. However, scalable development around these environments remains constrained by the absence of a systematic framework, especially one for synthesizing verifiable training data and integrating it with agent training and diagnostic evaluation. To address this challenge, we present ClawGym, a scalable framework that supports the full lifecycle of Claw-style personal agent development. Concretely, we construct ClawGym-SynData, a diverse dataset of 13.5K filtered tasks synthesized from pers"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ClawGym supports the full lifecycle of Claw-style personal agent development by constructing ClawGym-SynData (13.5K filtered tasks), training ClawGym-Agents via SFT on black-box rollouts plus lightweight RL, and providing ClawGym-Bench (200 instances) for reliable evaluation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The persona-driven synthetic tasks and hybrid verification mechanisms produce training data and evaluations that transfer to real-world Claw-style environments with persistent local state and external tools.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ClawGym supplies a 13.5K-task synthetic dataset, SFT-plus-RL trained agents, and a 200-instance benchmark to support the full lifecycle of Claw-style personal agent development.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ClawGym provides a complete framework for synthesizing data, training, and evaluating Claw-style personal agents.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0e492ca1be6dbeda081de9673c434b6da9d5729b5684342c91f3a2357c64c504"},"source":{"id":"2604.26904","kind":"arxiv","version":3},"verdict":{"id":"57d037e6-9cfe-4881-8cc9-50982bdb1bb6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T06:59:15.857554Z","strongest_claim":"ClawGym supports the full lifecycle of Claw-style personal agent development by constructing ClawGym-SynData (13.5K filtered tasks), training ClawGym-Agents via SFT on black-box rollouts plus lightweight RL, and providing ClawGym-Bench (200 instances) for reliable evaluation.","one_line_summary":"ClawGym supplies a 13.5K-task synthetic dataset, SFT-plus-RL trained agents, and a 200-instance benchmark to support the full lifecycle of Claw-style personal agent development.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The persona-driven synthetic tasks and hybrid verification mechanisms produce training data and evaluations that transfer to real-world Claw-style environments with persistent local state and external tools.","pith_extraction_headline":"ClawGym provides a complete framework for synthesizing data, training, and evaluating Claw-style personal agents."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.26904/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T19:43:24.121806Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"49f1459fbd9ab94b8bcd8d405836e803d43d53145cc177a949039bf49729e1fa"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}