{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:36SVPK76GLVA4V7VGTNPKRVUXK","short_pith_number":"pith:36SVPK76","schema_version":"1.0","canonical_sha256":"dfa557abfe32ea0e57f534daf546b4baba64b72d3227ae45c630b0d88b7c3285","source":{"kind":"arxiv","id":"2606.05625","version":1},"attestation_state":"computed","paper":{"title":"Self-Commitment Latency: A Reward-Free Probe for Prompted Implicit Hacking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Bonan Shen, Dingyan Shang, Tao Ning, Youting Wang","submitted_at":"2026-06-04T02:50:26Z","abstract_excerpt":"Implicit reward hacking is hard to audit when a language model's chain of thought appears benign: a final answer may be anchored by a prompt shortcut while the written reasoning still resembles ordinary problem solving. Verifier-based probes expose such behavior by measuring how early truncated reasoning contexts obtain high reward, but require a task-specific reward signal. This paper proposes a weaker-input alternative, self-commitment latency, which measures how early a prompted reasoning context commits to the model's own final answer. We evaluate the probe in a controlled paired GSM8K set"},"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":"2606.05625","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-06-04T02:50:26Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2326fbadebf0d98cd7912255319651bd143c4701043e7da0b0638d8ed582dd03","abstract_canon_sha256":"6d295b9d9912219d50c7cebde24bec56f0f3e570617bd2dc2792729fe93513c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:14:56.944753Z","signature_b64":"aWMwO3hZ0tmLLOLmUzw258EVaC1o4CV+imTARf4+O+J+P/fTgkispS+grwGl+W9Zw1NxCOxdjC7OL7Y+wCv5BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dfa557abfe32ea0e57f534daf546b4baba64b72d3227ae45c630b0d88b7c3285","last_reissued_at":"2026-06-05T01:14:56.944202Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:14:56.944202Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-Commitment Latency: A Reward-Free Probe for Prompted Implicit Hacking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Bonan Shen, Dingyan Shang, Tao Ning, Youting Wang","submitted_at":"2026-06-04T02:50:26Z","abstract_excerpt":"Implicit reward hacking is hard to audit when a language model's chain of thought appears benign: a final answer may be anchored by a prompt shortcut while the written reasoning still resembles ordinary problem solving. Verifier-based probes expose such behavior by measuring how early truncated reasoning contexts obtain high reward, but require a task-specific reward signal. This paper proposes a weaker-input alternative, self-commitment latency, which measures how early a prompted reasoning context commits to the model's own final answer. We evaluate the probe in a controlled paired GSM8K set"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.05625","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/2606.05625/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":"2606.05625","created_at":"2026-06-05T01:14:56.944287+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.05625v1","created_at":"2026-06-05T01:14:56.944287+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.05625","created_at":"2026-06-05T01:14:56.944287+00:00"},{"alias_kind":"pith_short_12","alias_value":"36SVPK76GLVA","created_at":"2026-06-05T01:14:56.944287+00:00"},{"alias_kind":"pith_short_16","alias_value":"36SVPK76GLVA4V7V","created_at":"2026-06-05T01:14:56.944287+00:00"},{"alias_kind":"pith_short_8","alias_value":"36SVPK76","created_at":"2026-06-05T01:14:56.944287+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/36SVPK76GLVA4V7VGTNPKRVUXK","json":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK.json","graph_json":"https://pith.science/api/pith-number/36SVPK76GLVA4V7VGTNPKRVUXK/graph.json","events_json":"https://pith.science/api/pith-number/36SVPK76GLVA4V7VGTNPKRVUXK/events.json","paper":"https://pith.science/paper/36SVPK76"},"agent_actions":{"view_html":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK","download_json":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK.json","view_paper":"https://pith.science/paper/36SVPK76","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.05625&json=true","fetch_graph":"https://pith.science/api/pith-number/36SVPK76GLVA4V7VGTNPKRVUXK/graph.json","fetch_events":"https://pith.science/api/pith-number/36SVPK76GLVA4V7VGTNPKRVUXK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK/action/storage_attestation","attest_author":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK/action/author_attestation","sign_citation":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK/action/citation_signature","submit_replication":"https://pith.science/pith/36SVPK76GLVA4V7VGTNPKRVUXK/action/replication_record"}},"created_at":"2026-06-05T01:14:56.944287+00:00","updated_at":"2026-06-05T01:14:56.944287+00:00"}