{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:46EPMDWDBYUN53NWAI4TUUNFGV","short_pith_number":"pith:46EPMDWD","schema_version":"1.0","canonical_sha256":"e788f60ec30e28deedb602393a51a535712c6dee31795373a15d16fb0f0e1877","source":{"kind":"arxiv","id":"2608.09745","version":1},"attestation_state":"computed","paper":{"title":"SR-OPSD: Self-Referenced On-Policy Self-Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Che Liu, Entong Li, Harrison Bo Hua Zhu, Huihang Liu, Kaiyu Li, Li Zeng, Wenxuan Yuan, Xiaoyuan Cheng, Yanlong Zhao, Zhuo Sun","submitted_at":"2026-08-10T15:40:07Z","abstract_excerpt":"On-policy self-distillation (OPSD) converts feedback into dense token-level supervision on trajectories generated by the policy to be optimized, providing a useful complement to reinforcement learning with sparse outcome rewards. However, the self-teacher policy used in OPSD is typically a stop-gradient or exponential-moving-average copy of the policy conditioned on additional context information, and thus co-evolves with both the student policy and its on-policy context distribution. Directly matching such a moving target with a fixed projection objective can lead to unstable optimization or "},"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":"2608.09745","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-10T15:40:07Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"f75b5af941746c85b196894abab4245d5b92932c4da99d3960b409dacc0d5fd5","abstract_canon_sha256":"902a2a074efa80f169de7a7b8723345e1349375a4a6e75021ed47ac754694d49"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T02:24:55.970336Z","signature_b64":"dIK8AttOHG2K8ka42KAINhKm6g205wVpSCGYgh4BwgZeGTSih2F/n23dKhbTep14UWO46dvd5G41yn5L2xvbCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e788f60ec30e28deedb602393a51a535712c6dee31795373a15d16fb0f0e1877","last_reissued_at":"2026-08-11T02:24:55.968604Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T02:24:55.968604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SR-OPSD: Self-Referenced On-Policy Self-Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Che Liu, Entong Li, Harrison Bo Hua Zhu, Huihang Liu, Kaiyu Li, Li Zeng, Wenxuan Yuan, Xiaoyuan Cheng, Yanlong Zhao, Zhuo Sun","submitted_at":"2026-08-10T15:40:07Z","abstract_excerpt":"On-policy self-distillation (OPSD) converts feedback into dense token-level supervision on trajectories generated by the policy to be optimized, providing a useful complement to reinforcement learning with sparse outcome rewards. However, the self-teacher policy used in OPSD is typically a stop-gradient or exponential-moving-average copy of the policy conditioned on additional context information, and thus co-evolves with both the student policy and its on-policy context distribution. Directly matching such a moving target with a fixed projection objective can lead to unstable optimization or "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.09745","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/2608.09745/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":"2608.09745","created_at":"2026-08-11T02:24:55.969283+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.09745v1","created_at":"2026-08-11T02:24:55.969283+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.09745","created_at":"2026-08-11T02:24:55.969283+00:00"},{"alias_kind":"pith_short_12","alias_value":"46EPMDWDBYUN","created_at":"2026-08-11T02:24:55.969283+00:00"},{"alias_kind":"pith_short_16","alias_value":"46EPMDWDBYUN53NW","created_at":"2026-08-11T02:24:55.969283+00:00"},{"alias_kind":"pith_short_8","alias_value":"46EPMDWD","created_at":"2026-08-11T02:24:55.969283+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/46EPMDWDBYUN53NWAI4TUUNFGV","json":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV.json","graph_json":"https://pith.science/api/pith-number/46EPMDWDBYUN53NWAI4TUUNFGV/graph.json","events_json":"https://pith.science/api/pith-number/46EPMDWDBYUN53NWAI4TUUNFGV/events.json","paper":"https://pith.science/paper/46EPMDWD"},"agent_actions":{"view_html":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV","download_json":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV.json","view_paper":"https://pith.science/paper/46EPMDWD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.09745&json=true","fetch_graph":"https://pith.science/api/pith-number/46EPMDWDBYUN53NWAI4TUUNFGV/graph.json","fetch_events":"https://pith.science/api/pith-number/46EPMDWDBYUN53NWAI4TUUNFGV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV/action/storage_attestation","attest_author":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV/action/author_attestation","sign_citation":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV/action/citation_signature","submit_replication":"https://pith.science/pith/46EPMDWDBYUN53NWAI4TUUNFGV/action/replication_record"}},"created_at":"2026-08-11T02:24:55.969283+00:00","updated_at":"2026-08-11T02:24:55.969283+00:00"}