{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:Q445V2DCVQ6M3KHXBJIGJW3OA7","short_pith_number":"pith:Q445V2DC","schema_version":"1.0","canonical_sha256":"8739dae862ac3ccda8f70a5064db6e07c39155e79232627416965fdcd8b6012f","source":{"kind":"arxiv","id":"2606.17321","version":1},"attestation_state":"computed","paper":{"title":"ProCUA-SFT Technical Report","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Amala Sanjay Deshmukh, Andrew Tao, Brandon Cui, Hao Zhang, Jaehun Jung, Jan Kautz, Jin Xu, Karan Sapra, Mingjie Liu, Muhammad Khalifa, Shaokun Zhang, Ximing Lu, Yejin Choi, Yi Dong","submitted_at":"2026-06-15T22:04:11Z","abstract_excerpt":"Training computer-use agents (CUAs) -- models that interact with graphical desktops through screenshots and keyboard/mouse actions -- requires large-scale, diverse trajectory data collected in full desktop environments. The largest public resource, AgentNet (22.5K human trajectories), leads to negative transfer when used for supervised fine-tuning (SFT): continuing training UI-TARS 7B on AgentNet causes OSWorld success rate to fall from 26.3% to 8-10%. We present ProCUA-SFT, a dataset of 3.1M step-level SFT samples distilled from 93K synthetic trajectories across 2,484 application combinations"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2606.17321","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-15T22:04:11Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"02a8838c7bc07cce5957b75f59336681a49964dda500040901113186413eedee","abstract_canon_sha256":"a3de021fff3e06d9d12dfce523b15f63c8b8cb9e6ac7c6fe09d7fc0d38650bdc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:10:09.157030Z","signature_b64":"nUNfAl/zXec3t1biCsNOKVXxq9bzlE35/ZdjxQYkTKoCCvI+4QLWke8tN2rNFq/8NnY0o/hMwmfEZd0yL+0FCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8739dae862ac3ccda8f70a5064db6e07c39155e79232627416965fdcd8b6012f","last_reissued_at":"2026-06-19T16:10:09.156676Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:10:09.156676Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ProCUA-SFT Technical Report","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Amala Sanjay Deshmukh, Andrew Tao, Brandon Cui, Hao Zhang, Jaehun Jung, Jan Kautz, Jin Xu, Karan Sapra, Mingjie Liu, Muhammad Khalifa, Shaokun Zhang, Ximing Lu, Yejin Choi, Yi Dong","submitted_at":"2026-06-15T22:04:11Z","abstract_excerpt":"Training computer-use agents (CUAs) -- models that interact with graphical desktops through screenshots and keyboard/mouse actions -- requires large-scale, diverse trajectory data collected in full desktop environments. The largest public resource, AgentNet (22.5K human trajectories), leads to negative transfer when used for supervised fine-tuning (SFT): continuing training UI-TARS 7B on AgentNet causes OSWorld success rate to fall from 26.3% to 8-10%. We present ProCUA-SFT, a dataset of 3.1M step-level SFT samples distilled from 93K synthetic trajectories across 2,484 application combinations"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.17321","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2606.17321/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2606.17321","created_at":"2026-06-19T16:10:09.156738+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.17321v1","created_at":"2026-06-19T16:10:09.156738+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.17321","created_at":"2026-06-19T16:10:09.156738+00:00"},{"alias_kind":"pith_short_12","alias_value":"Q445V2DCVQ6M","created_at":"2026-06-19T16:10:09.156738+00:00"},{"alias_kind":"pith_short_16","alias_value":"Q445V2DCVQ6M3KHX","created_at":"2026-06-19T16:10:09.156738+00:00"},{"alias_kind":"pith_short_8","alias_value":"Q445V2DC","created_at":"2026-06-19T16:10:09.156738+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7","json":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7.json","graph_json":"https://pith.science/api/pith-number/Q445V2DCVQ6M3KHXBJIGJW3OA7/graph.json","events_json":"https://pith.science/api/pith-number/Q445V2DCVQ6M3KHXBJIGJW3OA7/events.json","paper":"https://pith.science/paper/Q445V2DC"},"agent_actions":{"view_html":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7","download_json":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7.json","view_paper":"https://pith.science/paper/Q445V2DC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.17321&json=true","fetch_graph":"https://pith.science/api/pith-number/Q445V2DCVQ6M3KHXBJIGJW3OA7/graph.json","fetch_events":"https://pith.science/api/pith-number/Q445V2DCVQ6M3KHXBJIGJW3OA7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7/action/storage_attestation","attest_author":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7/action/author_attestation","sign_citation":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7/action/citation_signature","submit_replication":"https://pith.science/pith/Q445V2DCVQ6M3KHXBJIGJW3OA7/action/replication_record"}},"created_at":"2026-06-19T16:10:09.156738+00:00","updated_at":"2026-06-19T16:10:09.156738+00:00"}