{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:M4ZCIMTOEYJT6GSEBA6FLVY3ZJ","short_pith_number":"pith:M4ZCIMTO","schema_version":"1.0","canonical_sha256":"673224326e26133f1a44083c55d71bca7f86832afe205b797249ef50fe83b467","source":{"kind":"arxiv","id":"2606.08893","version":1},"attestation_state":"computed","paper":{"title":"Cheap Reward Hacking Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Iv\\'an Belenky, Joaqu\\'in Itria, Steven Johns","submitted_at":"2026-06-08T00:35:54Z","abstract_excerpt":"A small transformer encoder is trained to map Terminal-Wrench trajectories onto a unit sphere where embedding distance approximates the $L_1$ distance between reward and metadata signals. A linear probe on top of that embedding detects reward hacking on the cleaned test split with AUC $0.9467$ and TPR@5%FPR $0.8296$, matching the TW sanitized LLM-as-judge AUC ($0.9510$ on the cleaned split) and exceeding its TPR@5%FPR ($0.7130$ vs $0.8296$) on the same information condition, at roughly four orders of magnitude lower per-trajectory cost. The encoder is not a pure behavior reader: stripping natu"},"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":"2606.08893","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-08T00:35:54Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"6597ce8a4bea34fb770c04b01bd61d948464f00203abb3b7d5ce93b7070ed65d","abstract_canon_sha256":"576f9b213aa6a9fcc2bcdc82f786a92707af58bbc64a6fc42620e337fd963a02"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:07:45.656711Z","signature_b64":"iFNOe5BKmngw+PShvESMXkoa5Qc4qyWDs0xRRvj96kASP3qU+KrikuPFdXT7QdzG/MrI4cGV2ZHxL4/p2v5RDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"673224326e26133f1a44083c55d71bca7f86832afe205b797249ef50fe83b467","last_reissued_at":"2026-06-09T02:07:45.655852Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:07:45.655852Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cheap Reward Hacking Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Iv\\'an Belenky, Joaqu\\'in Itria, Steven Johns","submitted_at":"2026-06-08T00:35:54Z","abstract_excerpt":"A small transformer encoder is trained to map Terminal-Wrench trajectories onto a unit sphere where embedding distance approximates the $L_1$ distance between reward and metadata signals. A linear probe on top of that embedding detects reward hacking on the cleaned test split with AUC $0.9467$ and TPR@5%FPR $0.8296$, matching the TW sanitized LLM-as-judge AUC ($0.9510$ on the cleaned split) and exceeding its TPR@5%FPR ($0.7130$ vs $0.8296$) on the same information condition, at roughly four orders of magnitude lower per-trajectory cost. The encoder is not a pure behavior reader: stripping natu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08893","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/2606.08893/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":"2606.08893","created_at":"2026-06-09T02:07:45.655994+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08893v1","created_at":"2026-06-09T02:07:45.655994+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08893","created_at":"2026-06-09T02:07:45.655994+00:00"},{"alias_kind":"pith_short_12","alias_value":"M4ZCIMTOEYJT","created_at":"2026-06-09T02:07:45.655994+00:00"},{"alias_kind":"pith_short_16","alias_value":"M4ZCIMTOEYJT6GSE","created_at":"2026-06-09T02:07:45.655994+00:00"},{"alias_kind":"pith_short_8","alias_value":"M4ZCIMTO","created_at":"2026-06-09T02:07:45.655994+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/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ","json":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ.json","graph_json":"https://pith.science/api/pith-number/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/graph.json","events_json":"https://pith.science/api/pith-number/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/events.json","paper":"https://pith.science/paper/M4ZCIMTO"},"agent_actions":{"view_html":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ","download_json":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ.json","view_paper":"https://pith.science/paper/M4ZCIMTO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08893&json=true","fetch_graph":"https://pith.science/api/pith-number/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/graph.json","fetch_events":"https://pith.science/api/pith-number/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/action/storage_attestation","attest_author":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/action/author_attestation","sign_citation":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/action/citation_signature","submit_replication":"https://pith.science/pith/M4ZCIMTOEYJT6GSEBA6FLVY3ZJ/action/replication_record"}},"created_at":"2026-06-09T02:07:45.655994+00:00","updated_at":"2026-06-09T02:07:45.655994+00:00"}