{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:33FOMLXTLCQ2IXW67XYFOAXUYY","short_pith_number":"pith:33FOMLXT","schema_version":"1.0","canonical_sha256":"decae62ef358a1a45edefdf05702f4c61f1cec36ecc93aa87587f164e6f767c4","source":{"kind":"arxiv","id":"2608.09476","version":1},"attestation_state":"computed","paper":{"title":"ActBench: Self-Evolving Benchmark of Behavioral Safety in Cowork Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Cong Wang, Hongwei Yao, Jieling Chen, Kui Ren, Meihui Chen, Wangze Ni, Yiling He, Yiming Liu, Zikun Chen","submitted_at":"2026-08-10T11:45:03Z","abstract_excerpt":"Cowork agents may complete benign tasks while disclosing protected data, manipulating unauthorized state, invocate unauthorized API. We define behavioral safety and introduce ActBench, a self-evolving benchmark that evaluates such behavior risk from execution trajectories rather than final responses. Each case pairs a benign task with an adversarial variant that preserves its instruction, configuration, initial state, rating model, and trusted records while injecting a task-reachable payload. ActBench contains 600 cases from 213 scenarios, spanning 15 risk behaviors, six execution spaces, and "},"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":"2608.09476","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-08-10T11:45:03Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9d77e3e68132f173a281e06df366e0930ae22639386c2ee0967d8d914c726828","abstract_canon_sha256":"3246aa99596efbab510353a7014614e06ce2f25cc2998b18c412d1240cb4f857"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T02:24:18.217447Z","signature_b64":"COMA4bSlIBUflbemyQ+KuJsNjGNXOJwbTEGn9keMTeyfDC5Fvj/Lun715G0o9ukfszdxzBHF1n/F1uu2JP3TCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"decae62ef358a1a45edefdf05702f4c61f1cec36ecc93aa87587f164e6f767c4","last_reissued_at":"2026-08-11T02:24:18.215852Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T02:24:18.215852Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ActBench: Self-Evolving Benchmark of Behavioral Safety in Cowork Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CR","authors_text":"Cong Wang, Hongwei Yao, Jieling Chen, Kui Ren, Meihui Chen, Wangze Ni, Yiling He, Yiming Liu, Zikun Chen","submitted_at":"2026-08-10T11:45:03Z","abstract_excerpt":"Cowork agents may complete benign tasks while disclosing protected data, manipulating unauthorized state, invocate unauthorized API. We define behavioral safety and introduce ActBench, a self-evolving benchmark that evaluates such behavior risk from execution trajectories rather than final responses. Each case pairs a benign task with an adversarial variant that preserves its instruction, configuration, initial state, rating model, and trusted records while injecting a task-reachable payload. ActBench contains 600 cases from 213 scenarios, spanning 15 risk behaviors, six execution spaces, and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.09476","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/2608.09476/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":"2608.09476","created_at":"2026-08-11T02:24:18.216476+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.09476v1","created_at":"2026-08-11T02:24:18.216476+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.09476","created_at":"2026-08-11T02:24:18.216476+00:00"},{"alias_kind":"pith_short_12","alias_value":"33FOMLXTLCQ2","created_at":"2026-08-11T02:24:18.216476+00:00"},{"alias_kind":"pith_short_16","alias_value":"33FOMLXTLCQ2IXW6","created_at":"2026-08-11T02:24:18.216476+00:00"},{"alias_kind":"pith_short_8","alias_value":"33FOMLXT","created_at":"2026-08-11T02:24:18.216476+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/33FOMLXTLCQ2IXW67XYFOAXUYY","json":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY.json","graph_json":"https://pith.science/api/pith-number/33FOMLXTLCQ2IXW67XYFOAXUYY/graph.json","events_json":"https://pith.science/api/pith-number/33FOMLXTLCQ2IXW67XYFOAXUYY/events.json","paper":"https://pith.science/paper/33FOMLXT"},"agent_actions":{"view_html":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY","download_json":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY.json","view_paper":"https://pith.science/paper/33FOMLXT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.09476&json=true","fetch_graph":"https://pith.science/api/pith-number/33FOMLXTLCQ2IXW67XYFOAXUYY/graph.json","fetch_events":"https://pith.science/api/pith-number/33FOMLXTLCQ2IXW67XYFOAXUYY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY/action/storage_attestation","attest_author":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY/action/author_attestation","sign_citation":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY/action/citation_signature","submit_replication":"https://pith.science/pith/33FOMLXTLCQ2IXW67XYFOAXUYY/action/replication_record"}},"created_at":"2026-08-11T02:24:18.216476+00:00","updated_at":"2026-08-11T02:24:18.216476+00:00"}