{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:NJWZPV64IO7ZCTWWIDXSPTV7ZM","short_pith_number":"pith:NJWZPV64","canonical_record":{"source":{"id":"2605.28184","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T09:07:06Z","cross_cats_sorted":[],"title_canon_sha256":"6eb05ef761edb4033965c02e201fc0369be4ac1c365f34e90e07eafa9756d20a","abstract_canon_sha256":"a9d22f5560c428abbea72904c78f459e3524918514a2fc58dcc58cab40d27881"},"schema_version":"1.0"},"canonical_sha256":"6a6d97d7dc43bf914ed640ef27cebfcb1814d4ee17dd4408302b30ad9e6db307","source":{"kind":"arxiv","id":"2605.28184","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.28184","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"arxiv_version","alias_value":"2605.28184v1","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28184","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_12","alias_value":"NJWZPV64IO7Z","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_16","alias_value":"NJWZPV64IO7ZCTWW","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_8","alias_value":"NJWZPV64","created_at":"2026-05-28T01:05:01Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:NJWZPV64IO7ZCTWWIDXSPTV7ZM","target":"record","payload":{"canonical_record":{"source":{"id":"2605.28184","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T09:07:06Z","cross_cats_sorted":[],"title_canon_sha256":"6eb05ef761edb4033965c02e201fc0369be4ac1c365f34e90e07eafa9756d20a","abstract_canon_sha256":"a9d22f5560c428abbea72904c78f459e3524918514a2fc58dcc58cab40d27881"},"schema_version":"1.0"},"canonical_sha256":"6a6d97d7dc43bf914ed640ef27cebfcb1814d4ee17dd4408302b30ad9e6db307","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:05:01.586370Z","signature_b64":"9aUb6kMgfsDKPToqgqHqbzk+QDcPZddIMtuBoHGGQm41dWg9TeH96arUWa9WSuhO+mQVFbbPTn04QbKmCyU2AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6a6d97d7dc43bf914ed640ef27cebfcb1814d4ee17dd4408302b30ad9e6db307","last_reissued_at":"2026-05-28T01:05:01.585951Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:05:01.585951Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.28184","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-28T01:05:01Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"rnRD6StyRSlMlEtnVeITaaHK0vsPrRCgtH7FXVNZAw8mvuDNMTLipKzB1Nx/qN9eME6OiFiiMsI9PNrSTm5OBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T02:34:19.366972Z"},"content_sha256":"dc433891a7c72ce915e9f4e98ede18f7d241bab0943ffb078dc0fc3d9815038d","schema_version":"1.0","event_id":"sha256:dc433891a7c72ce915e9f4e98ede18f7d241bab0943ffb078dc0fc3d9815038d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:NJWZPV64IO7ZCTWWIDXSPTV7ZM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Joint Training of Multi-Token Prediction in Reinforcement Learning via Optimal Coefficient Calibration","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guojun Yin, Jiajun Chai, Lin Chen, Shiming Xiang, Xiaohan Wang, Zili Wang","submitted_at":"2026-05-27T09:07:06Z","abstract_excerpt":"Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as the standard paradigm for improving reasoning capability of large language models, while Multi-Token Prediction (MTP) has been a widely adopted module in pretraining. Combining them is a natural approach, yet current RL practices detach MTP gradients because joint training degrades the performance. We revisit this failure from an optimization perspective. We show that the per-step effect of MTP on the RL objective can be decomposed into two terms: a first-order correlation and a second-order perturbation penalty. This decompo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28184","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/2605.28184/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":"2026-05-28T01:05:01Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"oNTLdgIZYEFKooX8ysHcEEezJFBrWrQnzyEHCR/izHkR4bcCswitM1+vqGkL6xbwGlnd0kTzIacoK6CnxHaBDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T02:34:19.367724Z"},"content_sha256":"7256781e24bd4bc489bef82be581005592d1ce7efdb1f06c89e828443f53df2d","schema_version":"1.0","event_id":"sha256:7256781e24bd4bc489bef82be581005592d1ce7efdb1f06c89e828443f53df2d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/bundle.json","state_url":"https://pith.science/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/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":"2026-06-01T02:34:19Z","links":{"resolver":"https://pith.science/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM","bundle":"https://pith.science/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/bundle.json","state":"https://pith.science/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/NJWZPV64IO7ZCTWWIDXSPTV7ZM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:NJWZPV64IO7ZCTWWIDXSPTV7ZM","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":"a9d22f5560c428abbea72904c78f459e3524918514a2fc58dcc58cab40d27881","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T09:07:06Z","title_canon_sha256":"6eb05ef761edb4033965c02e201fc0369be4ac1c365f34e90e07eafa9756d20a"},"schema_version":"1.0","source":{"id":"2605.28184","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.28184","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"arxiv_version","alias_value":"2605.28184v1","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28184","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_12","alias_value":"NJWZPV64IO7Z","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_16","alias_value":"NJWZPV64IO7ZCTWW","created_at":"2026-05-28T01:05:01Z"},{"alias_kind":"pith_short_8","alias_value":"NJWZPV64","created_at":"2026-05-28T01:05:01Z"}],"graph_snapshots":[{"event_id":"sha256:7256781e24bd4bc489bef82be581005592d1ce7efdb1f06c89e828443f53df2d","target":"graph","created_at":"2026-05-28T01:05:01Z","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/2605.28184/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as the standard paradigm for improving reasoning capability of large language models, while Multi-Token Prediction (MTP) has been a widely adopted module in pretraining. Combining them is a natural approach, yet current RL practices detach MTP gradients because joint training degrades the performance. We revisit this failure from an optimization perspective. We show that the per-step effect of MTP on the RL objective can be decomposed into two terms: a first-order correlation and a second-order perturbation penalty. This decompo","authors_text":"Guojun Yin, Jiajun Chai, Lin Chen, Shiming Xiang, Xiaohan Wang, Zili Wang","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T09:07:06Z","title":"Joint Training of Multi-Token Prediction in Reinforcement Learning via Optimal Coefficient Calibration"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28184","kind":"arxiv","version":1},"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:dc433891a7c72ce915e9f4e98ede18f7d241bab0943ffb078dc0fc3d9815038d","target":"record","created_at":"2026-05-28T01:05:01Z","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":"a9d22f5560c428abbea72904c78f459e3524918514a2fc58dcc58cab40d27881","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T09:07:06Z","title_canon_sha256":"6eb05ef761edb4033965c02e201fc0369be4ac1c365f34e90e07eafa9756d20a"},"schema_version":"1.0","source":{"id":"2605.28184","kind":"arxiv","version":1}},"canonical_sha256":"6a6d97d7dc43bf914ed640ef27cebfcb1814d4ee17dd4408302b30ad9e6db307","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"6a6d97d7dc43bf914ed640ef27cebfcb1814d4ee17dd4408302b30ad9e6db307","first_computed_at":"2026-05-28T01:05:01.585951Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:05:01.585951Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"9aUb6kMgfsDKPToqgqHqbzk+QDcPZddIMtuBoHGGQm41dWg9TeH96arUWa9WSuhO+mQVFbbPTn04QbKmCyU2AA==","signature_status":"signed_v1","signed_at":"2026-05-28T01:05:01.586370Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.28184","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:dc433891a7c72ce915e9f4e98ede18f7d241bab0943ffb078dc0fc3d9815038d","sha256:7256781e24bd4bc489bef82be581005592d1ce7efdb1f06c89e828443f53df2d"],"state_sha256":"288fa63cfa5be038064a08e95ade9df229f847a0351179975ddfeb0f40c48d1b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"nIeBVQuhAUlSvLxfNntIlfX/yCDF9EbGCUqlrcjLgHeKSwdwWK3ftSSj0haW5AUqNu4HAnx7n2amzJlfix9EDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-01T02:34:19.371398Z","bundle_sha256":"788e0e1f8dce33c095381d9805caf3f57c0c35b5698f5e2a0d1b057e243926ec"}}