{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:4WDAUHA457DUW2WTSYB5FNEHEG","short_pith_number":"pith:4WDAUHA4","canonical_record":{"source":{"id":"2602.07078","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T03:16:04Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c554b01c096d1d3825197d7608432715ce7de78ed5eae3cdaa1cfa9a8615299a","abstract_canon_sha256":"5aaeead01686ab1b4c279cec546be484cd80a81db43f77b98d9bf6ed2bbda08f"},"schema_version":"1.0"},"canonical_sha256":"e5860a1c1cefc74b6ad39603d2b48721a989b33487196e8c602ffd747c0ba80b","source":{"kind":"arxiv","id":"2602.07078","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.07078","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"arxiv_version","alias_value":"2602.07078v2","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.07078","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_12","alias_value":"4WDAUHA457DU","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_16","alias_value":"4WDAUHA457DUW2WT","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_8","alias_value":"4WDAUHA4","created_at":"2026-06-23T02:13:20Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:4WDAUHA457DUW2WTSYB5FNEHEG","target":"record","payload":{"canonical_record":{"source":{"id":"2602.07078","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T03:16:04Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c554b01c096d1d3825197d7608432715ce7de78ed5eae3cdaa1cfa9a8615299a","abstract_canon_sha256":"5aaeead01686ab1b4c279cec546be484cd80a81db43f77b98d9bf6ed2bbda08f"},"schema_version":"1.0"},"canonical_sha256":"e5860a1c1cefc74b6ad39603d2b48721a989b33487196e8c602ffd747c0ba80b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T02:13:20.782073Z","signature_b64":"XagG2ZtjqCHj9K30nMzepMI2I8qjqDHG9gffiDeanGSZylXIVsP582QWFW9LXEds6fWwIdti6KPVUJSR1FepAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e5860a1c1cefc74b6ad39603d2b48721a989b33487196e8c602ffd747c0ba80b","last_reissued_at":"2026-06-23T02:13:20.781679Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T02:13:20.781679Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2602.07078","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-23T02:13:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TqDFU3LYbRufzMyWT+VwtWVUSaN41W6eyJVPynTgrv7cG7K6DpytPeBGn4Fd9GqtIq+zz0d/SCJl4NyNjw1OAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-02T19:38:43.596166Z"},"content_sha256":"977f1c2707df547f9187e23514f1116a7f1ee1de78e342ba3adec50727ae67fe","schema_version":"1.0","event_id":"sha256:977f1c2707df547f9187e23514f1116a7f1ee1de78e342ba3adec50727ae67fe"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:4WDAUHA457DUW2WTSYB5FNEHEG","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"The Optimal Token Baseline: Variance Reduction for Long-Horizon LLM-RL","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Baoxiang Wang, Ge Zhang, Jiacai Liu, Jiawei Xu, Longtao Zheng, Qian Liu, Tianle Cai, Wei Liu, Yaxiang Zhang, Yingru Li, Yuxuan Tong, Zhenghai Xue, Ziniu Li","submitted_at":"2026-02-06T03:16:04Z","abstract_excerpt":"Reinforcement Learning (RL) for Large Language Models (LLMs) often suffers from training collapse in long-horizon tasks due to exploding gradient variance. To mitigate this, a baseline is commonly introduced for advantage computation; however, traditional value models remain difficult to optimize, and standard group-based baselines overlook sequence heterogeneity. Although classic optimal baseline theory can achieve global variance reduction, it neglects token heterogeneity and requires prohibitive gradient-based computation. In this work, we derive the Optimal Token Baseline (OTB) from first "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.07078","kind":"arxiv","version":2},"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/2602.07078/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-23T02:13:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"zLEoC1By+IUAt78zNd2IGmLpM1DVDCJkhSaf/H1d8FvS8bRN4ny605jl8+6Ev5EupdEffNef9b8a5OZg51n4AA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-02T19:38:43.596577Z"},"content_sha256":"9331559155c4d990fa41d7cb70b4845b5f9a7b45f899bd28c5f84a930e21fd1d","schema_version":"1.0","event_id":"sha256:9331559155c4d990fa41d7cb70b4845b5f9a7b45f899bd28c5f84a930e21fd1d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/4WDAUHA457DUW2WTSYB5FNEHEG/bundle.json","state_url":"https://pith.science/pith/4WDAUHA457DUW2WTSYB5FNEHEG/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/4WDAUHA457DUW2WTSYB5FNEHEG/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-02T19:38:43Z","links":{"resolver":"https://pith.science/pith/4WDAUHA457DUW2WTSYB5FNEHEG","bundle":"https://pith.science/pith/4WDAUHA457DUW2WTSYB5FNEHEG/bundle.json","state":"https://pith.science/pith/4WDAUHA457DUW2WTSYB5FNEHEG/state.json","well_known_bundle":"https://pith.science/.well-known/pith/4WDAUHA457DUW2WTSYB5FNEHEG/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:4WDAUHA457DUW2WTSYB5FNEHEG","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":"5aaeead01686ab1b4c279cec546be484cd80a81db43f77b98d9bf6ed2bbda08f","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T03:16:04Z","title_canon_sha256":"c554b01c096d1d3825197d7608432715ce7de78ed5eae3cdaa1cfa9a8615299a"},"schema_version":"1.0","source":{"id":"2602.07078","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2602.07078","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"arxiv_version","alias_value":"2602.07078v2","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.07078","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_12","alias_value":"4WDAUHA457DU","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_16","alias_value":"4WDAUHA457DUW2WT","created_at":"2026-06-23T02:13:20Z"},{"alias_kind":"pith_short_8","alias_value":"4WDAUHA4","created_at":"2026-06-23T02:13:20Z"}],"graph_snapshots":[{"event_id":"sha256:9331559155c4d990fa41d7cb70b4845b5f9a7b45f899bd28c5f84a930e21fd1d","target":"graph","created_at":"2026-06-23T02:13:20Z","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/2602.07078/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement Learning (RL) for Large Language Models (LLMs) often suffers from training collapse in long-horizon tasks due to exploding gradient variance. To mitigate this, a baseline is commonly introduced for advantage computation; however, traditional value models remain difficult to optimize, and standard group-based baselines overlook sequence heterogeneity. Although classic optimal baseline theory can achieve global variance reduction, it neglects token heterogeneity and requires prohibitive gradient-based computation. In this work, we derive the Optimal Token Baseline (OTB) from first ","authors_text":"Baoxiang Wang, Ge Zhang, Jiacai Liu, Jiawei Xu, Longtao Zheng, Qian Liu, Tianle Cai, Wei Liu, Yaxiang Zhang, Yingru Li, Yuxuan Tong, Zhenghai Xue, Ziniu Li","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T03:16:04Z","title":"The Optimal Token Baseline: Variance Reduction for Long-Horizon LLM-RL"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.07078","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":"sha256:977f1c2707df547f9187e23514f1116a7f1ee1de78e342ba3adec50727ae67fe","target":"record","created_at":"2026-06-23T02:13:20Z","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":"5aaeead01686ab1b4c279cec546be484cd80a81db43f77b98d9bf6ed2bbda08f","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T03:16:04Z","title_canon_sha256":"c554b01c096d1d3825197d7608432715ce7de78ed5eae3cdaa1cfa9a8615299a"},"schema_version":"1.0","source":{"id":"2602.07078","kind":"arxiv","version":2}},"canonical_sha256":"e5860a1c1cefc74b6ad39603d2b48721a989b33487196e8c602ffd747c0ba80b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e5860a1c1cefc74b6ad39603d2b48721a989b33487196e8c602ffd747c0ba80b","first_computed_at":"2026-06-23T02:13:20.781679Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-23T02:13:20.781679Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XagG2ZtjqCHj9K30nMzepMI2I8qjqDHG9gffiDeanGSZylXIVsP582QWFW9LXEds6fWwIdti6KPVUJSR1FepAA==","signature_status":"signed_v1","signed_at":"2026-06-23T02:13:20.782073Z","signed_message":"canonical_sha256_bytes"},"source_id":"2602.07078","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:977f1c2707df547f9187e23514f1116a7f1ee1de78e342ba3adec50727ae67fe","sha256:9331559155c4d990fa41d7cb70b4845b5f9a7b45f899bd28c5f84a930e21fd1d"],"state_sha256":"62f293d2fa1a73194830d7cd3e3d0ddc561cbd537ba840db6a2edc85141e8b93"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+1sL250uzKrKau6PB3UH2PcLTMeoh8pwnlcRtdyBMq3iuakPd9a5EckI/r5TbtSMAFcOwXj8M7VX+86cjG/lBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-02T19:38:43.598653Z","bundle_sha256":"ee040e13922e7c964aa8384b9dd06c8dea79a7c5400edf6d3ad7e5bff83234c4"}}