{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FNKBILTQR3KMYOYTFHJKP6UNXA","short_pith_number":"pith:FNKBILTQ","schema_version":"1.0","canonical_sha256":"2b54142e708ed4cc3b1329d2a7fa8db80bfbddfd174df1f39e9ff090192ad5f1","source":{"kind":"arxiv","id":"2608.00301","version":1},"attestation_state":"computed","paper":{"title":"Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenyang Yu, Weida Zhao, Xujun Che, Yuchen Yuan","submitted_at":"2026-07-31T21:31:07Z","abstract_excerpt":"Error-penalized scoring rules ($+1$ for a correct answer, $-\\lambda$ for a wrong one, $0$ for abstaining) are increasingly prescribed against hallucination: a rational agent facing such a rule answers exactly when its correctness probability exceeds Chow's threshold $t^\\ast=\\lambda/(1+\\lambda)$. We prove that a KL-anchored gradient learner can do the opposite. When abstention is a discrete action, the reward gradient and the anchor's restoring force are throttled by the same gate-saturation factor and die together: under explicit conditions (among them, blanket answering loses score in expecta"},"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":"2608.00301","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-07-31T21:31:07Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0db57c36df70a20447f6d8038419df64d9f35ec6768c3bbe1fc06a09d3148309","abstract_canon_sha256":"2e2057b4249a1c94afa7a1f2c633fa9791c7a932e1774dbf97aa10faab0acbc0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T00:35:00.344603Z","signature_b64":"D2q1pcHJ38lI9kfIS5DbCXK8s0exXhnmPOsU4PXxX+Kts2tL2edKmifSVpDsqcoQZuuYuZt94RjRY0BLprl1Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2b54142e708ed4cc3b1329d2a7fa8db80bfbddfd174df1f39e9ff090192ad5f1","last_reissued_at":"2026-08-04T00:35:00.343178Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T00:35:00.343178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Abstention as an Action Can Kill Both the Reward Gradient and the KL Anchor: Collapse Law and Repair for Error-Penalized Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenyang Yu, Weida Zhao, Xujun Che, Yuchen Yuan","submitted_at":"2026-07-31T21:31:07Z","abstract_excerpt":"Error-penalized scoring rules ($+1$ for a correct answer, $-\\lambda$ for a wrong one, $0$ for abstaining) are increasingly prescribed against hallucination: a rational agent facing such a rule answers exactly when its correctness probability exceeds Chow's threshold $t^\\ast=\\lambda/(1+\\lambda)$. We prove that a KL-anchored gradient learner can do the opposite. When abstention is a discrete action, the reward gradient and the anchor's restoring force are throttled by the same gate-saturation factor and die together: under explicit conditions (among them, blanket answering loses score in expecta"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.00301","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/2608.00301/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":"2608.00301","created_at":"2026-08-04T00:35:00.344245+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.00301v1","created_at":"2026-08-04T00:35:00.344245+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.00301","created_at":"2026-08-04T00:35:00.344245+00:00"},{"alias_kind":"pith_short_12","alias_value":"FNKBILTQR3KM","created_at":"2026-08-04T00:35:00.344245+00:00"},{"alias_kind":"pith_short_16","alias_value":"FNKBILTQR3KMYOYT","created_at":"2026-08-04T00:35:00.344245+00:00"},{"alias_kind":"pith_short_8","alias_value":"FNKBILTQ","created_at":"2026-08-04T00:35:00.344245+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/FNKBILTQR3KMYOYTFHJKP6UNXA","json":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA.json","graph_json":"https://pith.science/api/pith-number/FNKBILTQR3KMYOYTFHJKP6UNXA/graph.json","events_json":"https://pith.science/api/pith-number/FNKBILTQR3KMYOYTFHJKP6UNXA/events.json","paper":"https://pith.science/paper/FNKBILTQ"},"agent_actions":{"view_html":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA","download_json":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA.json","view_paper":"https://pith.science/paper/FNKBILTQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.00301&json=true","fetch_graph":"https://pith.science/api/pith-number/FNKBILTQR3KMYOYTFHJKP6UNXA/graph.json","fetch_events":"https://pith.science/api/pith-number/FNKBILTQR3KMYOYTFHJKP6UNXA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA/action/storage_attestation","attest_author":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA/action/author_attestation","sign_citation":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA/action/citation_signature","submit_replication":"https://pith.science/pith/FNKBILTQR3KMYOYTFHJKP6UNXA/action/replication_record"}},"created_at":"2026-08-04T00:35:00.344245+00:00","updated_at":"2026-08-04T00:35:00.344245+00:00"}