{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XH2FD2HEQB5AWR3DMHLA4HHO5E","short_pith_number":"pith:XH2FD2HE","schema_version":"1.0","canonical_sha256":"b9f451e8e4807a0b476361d60e1ceee93aa28f106d97e2d41e88d9b2aa041dc0","source":{"kind":"arxiv","id":"2601.22136","version":2},"attestation_state":"computed","paper":{"title":"StepShield: When, Not Whether to Intervene on Rogue Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.SE"],"primary_cat":"cs.LG","authors_text":"Gloria Felicia (University of Virginia), Hemant Kumar (University of Arizona), Jinfeng He (Cornell University), Michael Eniolade (University of the Cumberlands), Milan Hussain Angati (California State University Northridge), Zitha Sasindran (Indian Institute of Science Bangalore)","submitted_at":"2026-01-29T18:55:46Z","abstract_excerpt":"Agent safety benchmarks measure whether a monitor detects harm, not when. Yet timing is the difference between intervention and autopsy. We introduce StepShield, the first benchmark that treats detection timeliness as a first-class metric. On 9,429 incident-grounded code-agent trajectories, we define the Early Intervention Rate (EIR): the fraction of detected rogue trajectories where the alert fires within a k-step window after the divergence point, isolating timing quality from coverage. This metric exposes what we call the Forensics Trap: a pattern-based guardrail with 847 rules achieves 86%"},"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":"2601.22136","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-01-29T18:55:46Z","cross_cats_sorted":["cs.AI","cs.CR","cs.SE"],"title_canon_sha256":"5ab9b7da36c07414d5a0d8c370efc553b5b8ad828b28ad124e202ac533c721ff","abstract_canon_sha256":"27982c3ff44bed23905eaa462a559466eab398056aa6bc8842ce18c4ee8b3742"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-08T01:18:24.978693Z","signature_b64":"67qag8d0BygslVOn7roaV4APWOyiJtP9pnrDV7YWvdk2hrM3ZKrFzCcM5K1/2gnM1n9V6BKfYIXM9pQMl5VrCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9f451e8e4807a0b476361d60e1ceee93aa28f106d97e2d41e88d9b2aa041dc0","last_reissued_at":"2026-07-08T01:18:24.978146Z","signature_status":"signed_v1","first_computed_at":"2026-07-08T01:18:24.978146Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"StepShield: When, Not Whether to Intervene on Rogue Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.SE"],"primary_cat":"cs.LG","authors_text":"Gloria Felicia (University of Virginia), Hemant Kumar (University of Arizona), Jinfeng He (Cornell University), Michael Eniolade (University of the Cumberlands), Milan Hussain Angati (California State University Northridge), Zitha Sasindran (Indian Institute of Science Bangalore)","submitted_at":"2026-01-29T18:55:46Z","abstract_excerpt":"Agent safety benchmarks measure whether a monitor detects harm, not when. Yet timing is the difference between intervention and autopsy. We introduce StepShield, the first benchmark that treats detection timeliness as a first-class metric. On 9,429 incident-grounded code-agent trajectories, we define the Early Intervention Rate (EIR): the fraction of detected rogue trajectories where the alert fires within a k-step window after the divergence point, isolating timing quality from coverage. This metric exposes what we call the Forensics Trap: a pattern-based guardrail with 847 rules achieves 86%"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.22136","kind":"arxiv","version":2},"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/2601.22136/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":"2601.22136","created_at":"2026-07-08T01:18:24.978217+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.22136v2","created_at":"2026-07-08T01:18:24.978217+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.22136","created_at":"2026-07-08T01:18:24.978217+00:00"},{"alias_kind":"pith_short_12","alias_value":"XH2FD2HEQB5A","created_at":"2026-07-08T01:18:24.978217+00:00"},{"alias_kind":"pith_short_16","alias_value":"XH2FD2HEQB5AWR3D","created_at":"2026-07-08T01:18:24.978217+00:00"},{"alias_kind":"pith_short_8","alias_value":"XH2FD2HE","created_at":"2026-07-08T01:18:24.978217+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.13631","citing_title":"ProjGuard: Safety Monitoring for Computer-Use Agents via Low-Dimensional Projections","ref_index":10,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E","json":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E.json","graph_json":"https://pith.science/api/pith-number/XH2FD2HEQB5AWR3DMHLA4HHO5E/graph.json","events_json":"https://pith.science/api/pith-number/XH2FD2HEQB5AWR3DMHLA4HHO5E/events.json","paper":"https://pith.science/paper/XH2FD2HE"},"agent_actions":{"view_html":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E","download_json":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E.json","view_paper":"https://pith.science/paper/XH2FD2HE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.22136&json=true","fetch_graph":"https://pith.science/api/pith-number/XH2FD2HEQB5AWR3DMHLA4HHO5E/graph.json","fetch_events":"https://pith.science/api/pith-number/XH2FD2HEQB5AWR3DMHLA4HHO5E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E/action/storage_attestation","attest_author":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E/action/author_attestation","sign_citation":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E/action/citation_signature","submit_replication":"https://pith.science/pith/XH2FD2HEQB5AWR3DMHLA4HHO5E/action/replication_record"}},"created_at":"2026-07-08T01:18:24.978217+00:00","updated_at":"2026-07-08T01:18:24.978217+00:00"}