{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TQX4TL2GWCWUQV46XPD7PVHPKA","short_pith_number":"pith:TQX4TL2G","schema_version":"1.0","canonical_sha256":"9c2fc9af46b0ad48579ebbc7f7d4ef50139e4146d4337560bdd134e29575b3e8","source":{"kind":"arxiv","id":"2605.22675","version":1},"attestation_state":"computed","paper":{"title":"Self-Policy Distillation via Capability-Selective Subspace Projection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangya Hao, Hanxue Liang, Yitong Shang, Yunbo Long, Zhuokai Zhao","submitted_at":"2026-05-21T16:18:41Z","abstract_excerpt":"Self-distillation bootstraps large language models (LLMs) by training on their own generations. However, existing methods either rely on external signals to curate self-generated outputs (e.g., correctness filtering, execution feedback, and reward search), which are costly and unavailable for the best-performing frontier models, or skip curation entirely and train on all raw outputs, an approach that is often domain-specific and hard to generalize. Both also share a deeper weakness that self-generated outputs entangle task-relevant capability with others, such as stylistic patterns, formatting"},"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":"2605.22675","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-21T16:18:41Z","cross_cats_sorted":[],"title_canon_sha256":"42f23c3ba0bfb4d0a2f42ce4f33a71d9636a667763d18a7fd418ab57ddb58958","abstract_canon_sha256":"b9797d9e29424c1e3d742dab9bd80b4c61469f0e463886a0fdcf007c1b9120f8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T02:04:49.637917Z","signature_b64":"hel5zUCZ5qjV8o8gZ63TS3pbQuXYOCPfCAJ/nwymYFKeT+n2a5sG2alwHOsi0yz8JvfWdYKW42IE2sQYqBRpDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9c2fc9af46b0ad48579ebbc7f7d4ef50139e4146d4337560bdd134e29575b3e8","last_reissued_at":"2026-05-22T02:04:49.637127Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T02:04:49.637127Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-Policy Distillation via Capability-Selective Subspace Projection","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangya Hao, Hanxue Liang, Yitong Shang, Yunbo Long, Zhuokai Zhao","submitted_at":"2026-05-21T16:18:41Z","abstract_excerpt":"Self-distillation bootstraps large language models (LLMs) by training on their own generations. However, existing methods either rely on external signals to curate self-generated outputs (e.g., correctness filtering, execution feedback, and reward search), which are costly and unavailable for the best-performing frontier models, or skip curation entirely and train on all raw outputs, an approach that is often domain-specific and hard to generalize. Both also share a deeper weakness that self-generated outputs entangle task-relevant capability with others, such as stylistic patterns, formatting"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22675","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/2605.22675/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":"2605.22675","created_at":"2026-05-22T02:04:49.637253+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22675v1","created_at":"2026-05-22T02:04:49.637253+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22675","created_at":"2026-05-22T02:04:49.637253+00:00"},{"alias_kind":"pith_short_12","alias_value":"TQX4TL2GWCWU","created_at":"2026-05-22T02:04:49.637253+00:00"},{"alias_kind":"pith_short_16","alias_value":"TQX4TL2GWCWUQV46","created_at":"2026-05-22T02:04:49.637253+00:00"},{"alias_kind":"pith_short_8","alias_value":"TQX4TL2G","created_at":"2026-05-22T02:04:49.637253+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/TQX4TL2GWCWUQV46XPD7PVHPKA","json":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA.json","graph_json":"https://pith.science/api/pith-number/TQX4TL2GWCWUQV46XPD7PVHPKA/graph.json","events_json":"https://pith.science/api/pith-number/TQX4TL2GWCWUQV46XPD7PVHPKA/events.json","paper":"https://pith.science/paper/TQX4TL2G"},"agent_actions":{"view_html":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA","download_json":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA.json","view_paper":"https://pith.science/paper/TQX4TL2G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22675&json=true","fetch_graph":"https://pith.science/api/pith-number/TQX4TL2GWCWUQV46XPD7PVHPKA/graph.json","fetch_events":"https://pith.science/api/pith-number/TQX4TL2GWCWUQV46XPD7PVHPKA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA/action/storage_attestation","attest_author":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA/action/author_attestation","sign_citation":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA/action/citation_signature","submit_replication":"https://pith.science/pith/TQX4TL2GWCWUQV46XPD7PVHPKA/action/replication_record"}},"created_at":"2026-05-22T02:04:49.637253+00:00","updated_at":"2026-05-22T02:04:49.637253+00:00"}