{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:WJBPH6XG6VAQD7ULFG3MQVQXHC","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f04a7db0ce9a8aaca46bb6cb8c9ce60a53c7335b72a9ccfd89fb80ac0af6e6b5","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-10-07T15:54:24Z","title_canon_sha256":"6fabb6417347564004a63aec21336baa57e57ec92199b3164180bdd9582977bf"},"schema_version":"1.0","source":{"id":"2510.06062","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2510.06062","created_at":"2026-05-20T00:00:25Z"},{"alias_kind":"arxiv_version","alias_value":"2510.06062v2","created_at":"2026-05-20T00:00:25Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.06062","created_at":"2026-05-20T00:00:25Z"},{"alias_kind":"pith_short_12","alias_value":"WJBPH6XG6VAQ","created_at":"2026-05-20T00:00:25Z"},{"alias_kind":"pith_short_16","alias_value":"WJBPH6XG6VAQD7UL","created_at":"2026-05-20T00:00:25Z"},{"alias_kind":"pith_short_8","alias_value":"WJBPH6XG","created_at":"2026-05-20T00:00:25Z"}],"graph_snapshots":[{"event_id":"sha256:9ea3b63e0fc72a12f728c15f248342a80f102a30b4730e770c124ad1be79ae32","target":"graph","created_at":"2026-05-20T00:00:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2510.06062/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning (RL) has shown great promise in large language models (LLMs) post-training, which typically rely on token-level clipping to maintain stability during optimization. Despite the empirical success of GRPO-style methods, we identify a fundamental and previously overlooked challenge in this popular Outcome-Supervised RL (OSRL) paradigm. We reveal that in OSRL, where advantages are shared across tokens within a response, importance sampling (IS) ratios deviate from their traditional purpose of distribution correction as in classic RL, which become token-level weights that allo","authors_text":"Fuzheng Zhang, Guorui Zhou, Jiakang Wang, Kun Gai, Lei Lin, Ling Pan, Qingpeng Cai, Runze Liu, Wenping Hu, Xiu Li","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-10-07T15:54:24Z","title":"When Importance Sampling Misallocates Credit: Asymmetric Ratios for Outcome-Supervised RL"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.06062","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:3706351280ddba6228e750e513ef334af432c9de6b8c353c9323f1610461b322","target":"record","created_at":"2026-05-20T00:00:25Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f04a7db0ce9a8aaca46bb6cb8c9ce60a53c7335b72a9ccfd89fb80ac0af6e6b5","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-10-07T15:54:24Z","title_canon_sha256":"6fabb6417347564004a63aec21336baa57e57ec92199b3164180bdd9582977bf"},"schema_version":"1.0","source":{"id":"2510.06062","kind":"arxiv","version":2}},"canonical_sha256":"b242f3fae6f54101fe8b29b6c85617388dcc106a4f81952f7980221ced6926a8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b242f3fae6f54101fe8b29b6c85617388dcc106a4f81952f7980221ced6926a8","first_computed_at":"2026-05-20T00:00:25.446716Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:25.446716Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"009DKiyTCpNcuriqVeM8BTUkKj/yn/unUrl3N97RU5lZUVSqikY6gntZ9ckl2vVA81Febmb8t7195WorgdiQBQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:25.447550Z","signed_message":"canonical_sha256_bytes"},"source_id":"2510.06062","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:3706351280ddba6228e750e513ef334af432c9de6b8c353c9323f1610461b322","sha256:9ea3b63e0fc72a12f728c15f248342a80f102a30b4730e770c124ad1be79ae32"],"state_sha256":"df10ddf8774d938668bc27db849e76c5743b800cd441749ea8e1e5bdac8baa47"}