{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5D7V4YGGRZJ55XM4MM3HQ4EPCK","short_pith_number":"pith:5D7V4YGG","schema_version":"1.0","canonical_sha256":"e8ff5e60c68e53dedd9c633678708f12a09686cefed995893efd820c8ea8ab8c","source":{"kind":"arxiv","id":"2605.21654","version":1},"attestation_state":"computed","paper":{"title":"Value-Gradient Hypothesis of RL for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Arip Asadulaev, Daniil Ognev, Karim Salta, Martin Takac","submitted_at":"2026-05-20T19:09:05Z","abstract_excerpt":"Reinforcement learning substantially improves pretrained language models, but it remains understudied why critic-free methods such as PPO and GRPO work as well as they do, and when they should provide the largest gains. We develop a value-gradient perspective of critic-free RL for LLM post-training. First, under a differentiable rollout and additive-noise parameterization, we show that the actor update is value-gradient-like in expectation: the backward pass propagates costates whose conditional expectation equals the value gradient. Second, for discrete transformer policies, we show that auto"},"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":"2605.21654","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T19:09:05Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"3f141cae6fec5a3aa47b0b1a523eb66e10e06cf6edca392e5a22f411cd1f3dab","abstract_canon_sha256":"c0368509d49130e999ec1b34eab365d03ecba6561008289a37f4f33a83a34f00"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:26.992331Z","signature_b64":"wrDKSmT83XCcaLK9+Z2THtoBt18IWTvKi5dOkm0nvZKE4359bXo/sMbd5i65rmal+zZMBLi6xRU+x1BJoNAIDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e8ff5e60c68e53dedd9c633678708f12a09686cefed995893efd820c8ea8ab8c","last_reissued_at":"2026-05-22T01:03:26.991607Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:26.991607Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Value-Gradient Hypothesis of RL for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Arip Asadulaev, Daniil Ognev, Karim Salta, Martin Takac","submitted_at":"2026-05-20T19:09:05Z","abstract_excerpt":"Reinforcement learning substantially improves pretrained language models, but it remains understudied why critic-free methods such as PPO and GRPO work as well as they do, and when they should provide the largest gains. We develop a value-gradient perspective of critic-free RL for LLM post-training. First, under a differentiable rollout and additive-noise parameterization, we show that the actor update is value-gradient-like in expectation: the backward pass propagates costates whose conditional expectation equals the value gradient. Second, for discrete transformer policies, we show that auto"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21654","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/2605.21654/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":"2605.21654","created_at":"2026-05-22T01:03:26.991729+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21654v1","created_at":"2026-05-22T01:03:26.991729+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21654","created_at":"2026-05-22T01:03:26.991729+00:00"},{"alias_kind":"pith_short_12","alias_value":"5D7V4YGGRZJ5","created_at":"2026-05-22T01:03:26.991729+00:00"},{"alias_kind":"pith_short_16","alias_value":"5D7V4YGGRZJ55XM4","created_at":"2026-05-22T01:03:26.991729+00:00"},{"alias_kind":"pith_short_8","alias_value":"5D7V4YGG","created_at":"2026-05-22T01:03:26.991729+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/5D7V4YGGRZJ55XM4MM3HQ4EPCK","json":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK.json","graph_json":"https://pith.science/api/pith-number/5D7V4YGGRZJ55XM4MM3HQ4EPCK/graph.json","events_json":"https://pith.science/api/pith-number/5D7V4YGGRZJ55XM4MM3HQ4EPCK/events.json","paper":"https://pith.science/paper/5D7V4YGG"},"agent_actions":{"view_html":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK","download_json":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK.json","view_paper":"https://pith.science/paper/5D7V4YGG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21654&json=true","fetch_graph":"https://pith.science/api/pith-number/5D7V4YGGRZJ55XM4MM3HQ4EPCK/graph.json","fetch_events":"https://pith.science/api/pith-number/5D7V4YGGRZJ55XM4MM3HQ4EPCK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK/action/storage_attestation","attest_author":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK/action/author_attestation","sign_citation":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK/action/citation_signature","submit_replication":"https://pith.science/pith/5D7V4YGGRZJ55XM4MM3HQ4EPCK/action/replication_record"}},"created_at":"2026-05-22T01:03:26.991729+00:00","updated_at":"2026-05-22T01:03:26.991729+00:00"}