{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ","short_pith_number":"pith:VCFS2ZGN","schema_version":"1.0","canonical_sha256":"a88b2d64cd288dc1780a86a725af3b3e67e315d646b330eb4d17b84cdc1c2844","source":{"kind":"arxiv","id":"2601.21476","version":2},"attestation_state":"computed","paper":{"title":"Thinking Seeds: Leveraging Historical Diversity for Position-Aware RL in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenxi Sun, Deyi Xiong, Lei Yang, Renren Jin, Wei Bi","submitted_at":"2026-01-29T09:56:15Z","abstract_excerpt":"On-policy reinforcement learning (RL) for language model post-training suffers from a fundamental tension: as training progresses, policy entropy collapses and sampling diversity diminishes, causing the model to ``forget'' its own earlier exploratory capacity. While off-policy data can restore diversity, existing methods mix entire trajectories at the sequence level, introducing severe policy mismatch and training instability. We argue that the core question is not \\emph{whether} to use off-policy data, but \\emph{where} in the sequence it should appear. Based on this insight, we propose \\textb"},"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":"2601.21476","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-01-29T09:56:15Z","cross_cats_sorted":[],"title_canon_sha256":"0cb3c6102e1868607be38be2de785ac4c95a26dda53a60e6547bbb2a35429043","abstract_canon_sha256":"d3e1f20fb786842b4ae00af71f121ed8b762d2534c2ec10f3153a216763ac812"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:19:51.561268Z","signature_b64":"rgVW3vlhIVE2PVl3uSDCzxptS5Nt07E4t1OAdcapGEHkL78udUtkFomacj91wQz0yjvFetHK0mZs4+xrx3tMBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a88b2d64cd288dc1780a86a725af3b3e67e315d646b330eb4d17b84cdc1c2844","last_reissued_at":"2026-07-09T01:19:51.560813Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:19:51.560813Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Thinking Seeds: Leveraging Historical Diversity for Position-Aware RL in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chenxi Sun, Deyi Xiong, Lei Yang, Renren Jin, Wei Bi","submitted_at":"2026-01-29T09:56:15Z","abstract_excerpt":"On-policy reinforcement learning (RL) for language model post-training suffers from a fundamental tension: as training progresses, policy entropy collapses and sampling diversity diminishes, causing the model to ``forget'' its own earlier exploratory capacity. While off-policy data can restore diversity, existing methods mix entire trajectories at the sequence level, introducing severe policy mismatch and training instability. We argue that the core question is not \\emph{whether} to use off-policy data, but \\emph{where} in the sequence it should appear. Based on this insight, we propose \\textb"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2601.21476","kind":"arxiv","version":2},"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/2601.21476/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":"2601.21476","created_at":"2026-07-09T01:19:51.560872+00:00"},{"alias_kind":"arxiv_version","alias_value":"2601.21476v2","created_at":"2026-07-09T01:19:51.560872+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2601.21476","created_at":"2026-07-09T01:19:51.560872+00:00"},{"alias_kind":"pith_short_12","alias_value":"VCFS2ZGNFCG4","created_at":"2026-07-09T01:19:51.560872+00:00"},{"alias_kind":"pith_short_16","alias_value":"VCFS2ZGNFCG4C6AK","created_at":"2026-07-09T01:19:51.560872+00:00"},{"alias_kind":"pith_short_8","alias_value":"VCFS2ZGN","created_at":"2026-07-09T01:19:51.560872+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/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ","json":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ.json","graph_json":"https://pith.science/api/pith-number/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/graph.json","events_json":"https://pith.science/api/pith-number/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/events.json","paper":"https://pith.science/paper/VCFS2ZGN"},"agent_actions":{"view_html":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ","download_json":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ.json","view_paper":"https://pith.science/paper/VCFS2ZGN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2601.21476&json=true","fetch_graph":"https://pith.science/api/pith-number/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/graph.json","fetch_events":"https://pith.science/api/pith-number/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/action/storage_attestation","attest_author":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/action/author_attestation","sign_citation":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/action/citation_signature","submit_replication":"https://pith.science/pith/VCFS2ZGNFCG4C6AKQ2TSLLZ3HZ/action/replication_record"}},"created_at":"2026-07-09T01:19:51.560872+00:00","updated_at":"2026-07-09T01:19:51.560872+00:00"}