{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:D3ZMZJXCA5OSSOAYNUDWLBHXLT","short_pith_number":"pith:D3ZMZJXC","schema_version":"1.0","canonical_sha256":"1ef2cca6e2075d2938186d076584f75cf3f72b2a20a1ebe05709d7fff991348b","source":{"kind":"arxiv","id":"2605.29782","version":1},"attestation_state":"computed","paper":{"title":"Hista and Numca: Estimate State Value Effectively for LLM Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Garry Yang, James Cheng, Jiqian Dong, Yizhou Tian, Yongqiang Chen, Zhitang Chen, Zizhe Chen","submitted_at":"2026-05-28T11:31:13Z","abstract_excerpt":"Reinforcement learning (RL) refines large language models (LLMs) by directly optimizing model behavior through reward signals. While accurate state value estimation is critical for stable training in classical RL, it remains an underexplored challenge in LLM post-training. In this work, we introduce the State Value Estimation Benchmark (SVEB) to assess state estimation within existing RL frameworks and show that critics in standard approaches like PPO collapse to a coarse group-average baseline. To address this, we propose two techniques: Numca, which leverages numerical spans as gradable mile"},"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":"2605.29782","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-28T11:31:13Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"3ce94661979d332008ae35d8b5f7a30acf800eb4fdc3c2fcf5fbdb047eaad636","abstract_canon_sha256":"eeec2f0e382b83748e19984188d683828f6d32e6379779d35fdf44a818bb3df8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-29T02:05:52.082611Z","signature_b64":"Vl7qRi4QutTzOyHxnMgD90TYE5OFPpjxX3fU+MGi08+cG2LCX7cSfOxohjoDwJIK9BvZW23iQrdC0Rg2Sup0CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ef2cca6e2075d2938186d076584f75cf3f72b2a20a1ebe05709d7fff991348b","last_reissued_at":"2026-05-29T02:05:52.081886Z","signature_status":"signed_v1","first_computed_at":"2026-05-29T02:05:52.081886Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hista and Numca: Estimate State Value Effectively for LLM Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Garry Yang, James Cheng, Jiqian Dong, Yizhou Tian, Yongqiang Chen, Zhitang Chen, Zizhe Chen","submitted_at":"2026-05-28T11:31:13Z","abstract_excerpt":"Reinforcement learning (RL) refines large language models (LLMs) by directly optimizing model behavior through reward signals. While accurate state value estimation is critical for stable training in classical RL, it remains an underexplored challenge in LLM post-training. In this work, we introduce the State Value Estimation Benchmark (SVEB) to assess state estimation within existing RL frameworks and show that critics in standard approaches like PPO collapse to a coarse group-average baseline. To address this, we propose two techniques: Numca, which leverages numerical spans as gradable mile"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.29782","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/2605.29782/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":"2605.29782","created_at":"2026-05-29T02:05:52.081989+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.29782v1","created_at":"2026-05-29T02:05:52.081989+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.29782","created_at":"2026-05-29T02:05:52.081989+00:00"},{"alias_kind":"pith_short_12","alias_value":"D3ZMZJXCA5OS","created_at":"2026-05-29T02:05:52.081989+00:00"},{"alias_kind":"pith_short_16","alias_value":"D3ZMZJXCA5OSSOAY","created_at":"2026-05-29T02:05:52.081989+00:00"},{"alias_kind":"pith_short_8","alias_value":"D3ZMZJXC","created_at":"2026-05-29T02:05:52.081989+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/D3ZMZJXCA5OSSOAYNUDWLBHXLT","json":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT.json","graph_json":"https://pith.science/api/pith-number/D3ZMZJXCA5OSSOAYNUDWLBHXLT/graph.json","events_json":"https://pith.science/api/pith-number/D3ZMZJXCA5OSSOAYNUDWLBHXLT/events.json","paper":"https://pith.science/paper/D3ZMZJXC"},"agent_actions":{"view_html":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT","download_json":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT.json","view_paper":"https://pith.science/paper/D3ZMZJXC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.29782&json=true","fetch_graph":"https://pith.science/api/pith-number/D3ZMZJXCA5OSSOAYNUDWLBHXLT/graph.json","fetch_events":"https://pith.science/api/pith-number/D3ZMZJXCA5OSSOAYNUDWLBHXLT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT/action/storage_attestation","attest_author":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT/action/author_attestation","sign_citation":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT/action/citation_signature","submit_replication":"https://pith.science/pith/D3ZMZJXCA5OSSOAYNUDWLBHXLT/action/replication_record"}},"created_at":"2026-05-29T02:05:52.081989+00:00","updated_at":"2026-05-29T02:05:52.081989+00:00"}