{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UDCHASBKWKWS2KPU6U37XJSCGO","short_pith_number":"pith:UDCHASBK","schema_version":"1.0","canonical_sha256":"a0c470482ab2ad2d29f4f537fba64233897083614f134179cfc38ce9e0c48525","source":{"kind":"arxiv","id":"2605.24998","version":1},"attestation_state":"computed","paper":{"title":"Better, Faster: Harnessing Self-Improvement in Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bo Du, Dacheng Tao, Juhua Liu, Leszek Rutkowski, Liang Ding, Qihuang Zhong","submitted_at":"2026-05-24T10:54:46Z","abstract_excerpt":"Self-improvement training enables the large reasoning models (LRMs) to improve themselves by self-generating reasoning trajectories as training data without external supervision. However, we find that this method often falls short in complex reasoning tasks and even leads to model collapse. Through a series of preliminary analyses, we reveal two problems: (1) data imbalance, where most training samples are simple, but the challenging yet crucial samples are scarce; (2) overthinking, where many undesired samples with redundant reasoning steps are used for self-training. To this end, we propose "},"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":"2605.24998","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-24T10:54:46Z","cross_cats_sorted":[],"title_canon_sha256":"c24f7d718356e4f4cb71afc28037869d67e3aa4ea1762ff1abe07d7227bfaf5d","abstract_canon_sha256":"c8e2e66a52a27878649b834cc716c4a6a7d41a6f3f3e6399305e894d67bd7551"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:04:09.889595Z","signature_b64":"F2wfsSnqPzL0SUn7/NMzUq7T9tvTn4WkRkfgCS+S4JsE8iLrbHvExyOAneoEWQmbAjVDlYPrTyrUvbjYaenwDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a0c470482ab2ad2d29f4f537fba64233897083614f134179cfc38ce9e0c48525","last_reissued_at":"2026-05-26T01:04:09.888821Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:04:09.888821Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Better, Faster: Harnessing Self-Improvement in Large Reasoning Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Bo Du, Dacheng Tao, Juhua Liu, Leszek Rutkowski, Liang Ding, Qihuang Zhong","submitted_at":"2026-05-24T10:54:46Z","abstract_excerpt":"Self-improvement training enables the large reasoning models (LRMs) to improve themselves by self-generating reasoning trajectories as training data without external supervision. However, we find that this method often falls short in complex reasoning tasks and even leads to model collapse. Through a series of preliminary analyses, we reveal two problems: (1) data imbalance, where most training samples are simple, but the challenging yet crucial samples are scarce; (2) overthinking, where many undesired samples with redundant reasoning steps are used for self-training. To this end, we propose "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.24998","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/2605.24998/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":"2605.24998","created_at":"2026-05-26T01:04:09.888910+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.24998v1","created_at":"2026-05-26T01:04:09.888910+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.24998","created_at":"2026-05-26T01:04:09.888910+00:00"},{"alias_kind":"pith_short_12","alias_value":"UDCHASBKWKWS","created_at":"2026-05-26T01:04:09.888910+00:00"},{"alias_kind":"pith_short_16","alias_value":"UDCHASBKWKWS2KPU","created_at":"2026-05-26T01:04:09.888910+00:00"},{"alias_kind":"pith_short_8","alias_value":"UDCHASBK","created_at":"2026-05-26T01:04:09.888910+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/UDCHASBKWKWS2KPU6U37XJSCGO","json":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO.json","graph_json":"https://pith.science/api/pith-number/UDCHASBKWKWS2KPU6U37XJSCGO/graph.json","events_json":"https://pith.science/api/pith-number/UDCHASBKWKWS2KPU6U37XJSCGO/events.json","paper":"https://pith.science/paper/UDCHASBK"},"agent_actions":{"view_html":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO","download_json":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO.json","view_paper":"https://pith.science/paper/UDCHASBK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.24998&json=true","fetch_graph":"https://pith.science/api/pith-number/UDCHASBKWKWS2KPU6U37XJSCGO/graph.json","fetch_events":"https://pith.science/api/pith-number/UDCHASBKWKWS2KPU6U37XJSCGO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO/action/storage_attestation","attest_author":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO/action/author_attestation","sign_citation":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO/action/citation_signature","submit_replication":"https://pith.science/pith/UDCHASBKWKWS2KPU6U37XJSCGO/action/replication_record"}},"created_at":"2026-05-26T01:04:09.888910+00:00","updated_at":"2026-05-26T01:04:09.888910+00:00"}