{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DZCIQQEJNW6VB6AUWBYNADQEG2","short_pith_number":"pith:DZCIQQEJ","schema_version":"1.0","canonical_sha256":"1e448840896dbd50f814b070d00e0436921df6961afb250effe8345c9aeac6bd","source":{"kind":"arxiv","id":"2606.01075","version":1},"attestation_state":"computed","paper":{"title":"On the Generalization Gap in Self-Evolving Language Model Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andrew Tomkins, Cyrus Rashtchian, Da-Cheng Juan, Kan Yuan, Stefanie Anna Baby, Susanna Maria Baby, Tu Vu, Zhenting Qi","submitted_at":"2026-05-31T07:43:19Z","abstract_excerpt":"Recent work suggests that large language models (LLMs) can improve through self-evolution (SE), using supervision signals generated by the model itself. In this work, we ask: under a strict closed-loop setup, where the self-evolution algorithm has access only to an unlabeled prompt set and a base model, how close can internally generated supervision come to oracle-supervised training? We analyze four representative strategies in a unified offline self-evolution framework: single-round verification, multi-turn revision with feedback, iterative training, and curriculum learning. Our primary expe"},"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":"2606.01075","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-31T07:43:19Z","cross_cats_sorted":[],"title_canon_sha256":"3dc0acf149b962945cec8ecc6aa1b44de65e93a77e9e249828ceeb04715b8a7d","abstract_canon_sha256":"aaae93f5f001b7d323572c0afb3761fbdba7f9afcba5ba6de5f49ebf51f62d1e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:04:20.595984Z","signature_b64":"Cp7GGTRjMeU27q/gQlaEpLD4rR9h4YAn1HTsCON2uwBbkiYWj9DCNua0jGKUMnAgmXZxukJ05/b7gyLiVnPLDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1e448840896dbd50f814b070d00e0436921df6961afb250effe8345c9aeac6bd","last_reissued_at":"2026-06-02T01:04:20.595559Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:04:20.595559Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Generalization Gap in Self-Evolving Language Model Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Andrew Tomkins, Cyrus Rashtchian, Da-Cheng Juan, Kan Yuan, Stefanie Anna Baby, Susanna Maria Baby, Tu Vu, Zhenting Qi","submitted_at":"2026-05-31T07:43:19Z","abstract_excerpt":"Recent work suggests that large language models (LLMs) can improve through self-evolution (SE), using supervision signals generated by the model itself. In this work, we ask: under a strict closed-loop setup, where the self-evolution algorithm has access only to an unlabeled prompt set and a base model, how close can internally generated supervision come to oracle-supervised training? We analyze four representative strategies in a unified offline self-evolution framework: single-round verification, multi-turn revision with feedback, iterative training, and curriculum learning. Our primary expe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.01075","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/2606.01075/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":"2606.01075","created_at":"2026-06-02T01:04:20.595618+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.01075v1","created_at":"2026-06-02T01:04:20.595618+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.01075","created_at":"2026-06-02T01:04:20.595618+00:00"},{"alias_kind":"pith_short_12","alias_value":"DZCIQQEJNW6V","created_at":"2026-06-02T01:04:20.595618+00:00"},{"alias_kind":"pith_short_16","alias_value":"DZCIQQEJNW6VB6AU","created_at":"2026-06-02T01:04:20.595618+00:00"},{"alias_kind":"pith_short_8","alias_value":"DZCIQQEJ","created_at":"2026-06-02T01:04:20.595618+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/DZCIQQEJNW6VB6AUWBYNADQEG2","json":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2.json","graph_json":"https://pith.science/api/pith-number/DZCIQQEJNW6VB6AUWBYNADQEG2/graph.json","events_json":"https://pith.science/api/pith-number/DZCIQQEJNW6VB6AUWBYNADQEG2/events.json","paper":"https://pith.science/paper/DZCIQQEJ"},"agent_actions":{"view_html":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2","download_json":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2.json","view_paper":"https://pith.science/paper/DZCIQQEJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.01075&json=true","fetch_graph":"https://pith.science/api/pith-number/DZCIQQEJNW6VB6AUWBYNADQEG2/graph.json","fetch_events":"https://pith.science/api/pith-number/DZCIQQEJNW6VB6AUWBYNADQEG2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2/action/storage_attestation","attest_author":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2/action/author_attestation","sign_citation":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2/action/citation_signature","submit_replication":"https://pith.science/pith/DZCIQQEJNW6VB6AUWBYNADQEG2/action/replication_record"}},"created_at":"2026-06-02T01:04:20.595618+00:00","updated_at":"2026-06-02T01:04:20.595618+00:00"}