{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:RXMOX2BKGDLE6QNAPS4KDJJSHA","short_pith_number":"pith:RXMOX2BK","canonical_record":{"source":{"id":"2604.13088","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-04T02:47:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"46d7e74857a28d4a18ccaf7d2c9667372fb0b4eaa2388ddd2b52c1205e6d697b","abstract_canon_sha256":"7a24e707f83e0a91d08fdb7046463ad76317ecd108b7c3bcabac36b363346fff"},"schema_version":"1.0"},"canonical_sha256":"8dd8ebe82a30d64f41a07cb8a1a532380dc2694bcd29f380244ef81f424422b5","source":{"kind":"arxiv","id":"2604.13088","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.13088","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"arxiv_version","alias_value":"2604.13088v2","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.13088","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_12","alias_value":"RXMOX2BKGDLE","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_16","alias_value":"RXMOX2BKGDLE6QNA","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_8","alias_value":"RXMOX2BK","created_at":"2026-05-26T01:02:34Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:RXMOX2BKGDLE6QNAPS4KDJJSHA","target":"record","payload":{"canonical_record":{"source":{"id":"2604.13088","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-04T02:47:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"46d7e74857a28d4a18ccaf7d2c9667372fb0b4eaa2388ddd2b52c1205e6d697b","abstract_canon_sha256":"7a24e707f83e0a91d08fdb7046463ad76317ecd108b7c3bcabac36b363346fff"},"schema_version":"1.0"},"canonical_sha256":"8dd8ebe82a30d64f41a07cb8a1a532380dc2694bcd29f380244ef81f424422b5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:02:34.141505Z","signature_b64":"3WbpDGPf8koi5z0Ph1GP3X8R+fLZODojXrqimzWbLdz3CdG+wRv1sssa5e4upjbPBvO60X5QmxRsMoZZnhtxDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8dd8ebe82a30d64f41a07cb8a1a532380dc2694bcd29f380244ef81f424422b5","last_reissued_at":"2026-05-26T01:02:34.140627Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:02:34.140627Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.13088","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":"2026-05-26T01:02:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"nOdih29yAnraVFwL0V3lzJ16lL3G/iRZxE7tRsKwavmTDY8GvRGVNM5/BaS3nPiu5VpQiLu190fqC1A80jG7Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T04:49:07.573344Z"},"content_sha256":"dda5fcecc740f25bf544e3d8b75e8b3212e2b41b6bc2b6abc011fa2f06ddeaad","schema_version":"1.0","event_id":"sha256:dda5fcecc740f25bf544e3d8b75e8b3212e2b41b6bc2b6abc011fa2f06ddeaad"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:RXMOX2BKGDLE6QNAPS4KDJJSHA","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Design Conditions for Intra-Group Learning of Sequence-Level Rewards: Token Gradient Cancellation","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Fei Ding, Yongkang Zhang, youwei wang, Zijian Zeng","submitted_at":"2026-04-04T02:47:28Z","abstract_excerpt":"Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"to prevent reward-irrelevant drift, intra-group objectives must maintain gradient exchangeability across token updates, enabling gradient cancellation on weak-credit/high-frequency tokens.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the observed training failures (learning tax, solution probability drift, entropy collapse) are primarily caused by disruption of token-level gradient exchangeability rather than other factors such as reward sparsity itself or optimizer dynamics.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Intra-group objectives in sparse-reward RL must maintain token gradient exchangeability to enable cancellation on weak-credit tokens and stabilize training.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3f85b0770105bfd8309f1126dc1cc79601f3b00ab70d3d0a17db89cea286f84c"},"source":{"id":"2604.13088","kind":"arxiv","version":2},"verdict":{"id":"b38de77e-6f47-426b-9868-4fb6a5965e2a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T18:38:00.557651Z","strongest_claim":"to prevent reward-irrelevant drift, intra-group objectives must maintain gradient exchangeability across token updates, enabling gradient cancellation on weak-credit/high-frequency tokens.","one_line_summary":"Intra-group objectives in sparse-reward RL must maintain token gradient exchangeability to enable cancellation on weak-credit tokens and stabilize training.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the observed training failures (learning tax, solution probability drift, entropy collapse) are primarily caused by disruption of token-level gradient exchangeability rather than other factors such as reward sparsity itself or optimizer dynamics.","pith_extraction_headline":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.13088/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":1,"snapshot_sha256":"a7ec499fa2984d6b9f2c116cd0ee6128a352566b47d42486cde3d12b97a0a9c0"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"b38de77e-6f47-426b-9868-4fb6a5965e2a"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T01:02:34Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Lry3qieJe1R/sYHyukea/TmhYeSBPtsLReqikVWLZArdzDmv/61BkFJXDdfInGxFCDsuHAlI3GBarLOzSNSsDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-10T04:49:07.574259Z"},"content_sha256":"beddecf6754261519eccb5fdaff0cebf6a9c4120f77569a7dd370a337a01bbbd","schema_version":"1.0","event_id":"sha256:beddecf6754261519eccb5fdaff0cebf6a9c4120f77569a7dd370a337a01bbbd"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/bundle.json","state_url":"https://pith.science/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/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":"2026-06-10T04:49:07Z","links":{"resolver":"https://pith.science/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA","bundle":"https://pith.science/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/bundle.json","state":"https://pith.science/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RXMOX2BKGDLE6QNAPS4KDJJSHA/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:RXMOX2BKGDLE6QNAPS4KDJJSHA","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":"7a24e707f83e0a91d08fdb7046463ad76317ecd108b7c3bcabac36b363346fff","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-04T02:47:28Z","title_canon_sha256":"46d7e74857a28d4a18ccaf7d2c9667372fb0b4eaa2388ddd2b52c1205e6d697b"},"schema_version":"1.0","source":{"id":"2604.13088","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.13088","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"arxiv_version","alias_value":"2604.13088v2","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.13088","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_12","alias_value":"RXMOX2BKGDLE","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_16","alias_value":"RXMOX2BKGDLE6QNA","created_at":"2026-05-26T01:02:34Z"},{"alias_kind":"pith_short_8","alias_value":"RXMOX2BK","created_at":"2026-05-26T01:02:34Z"}],"graph_snapshots":[{"event_id":"sha256:beddecf6754261519eccb5fdaff0cebf6a9c4120f77569a7dd370a337a01bbbd","target":"graph","created_at":"2026-05-26T01:02:34Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"to prevent reward-irrelevant drift, intra-group objectives must maintain gradient exchangeability across token updates, enabling gradient cancellation on weak-credit/high-frequency tokens."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the observed training failures (learning tax, solution probability drift, entropy collapse) are primarily caused by disruption of token-level gradient exchangeability rather than other factors such as reward sparsity itself or optimizer dynamics."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Intra-group objectives in sparse-reward RL must maintain token gradient exchangeability to enable cancellation on weak-credit tokens and stabilize training."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift."}],"snapshot_sha256":"3f85b0770105bfd8309f1126dc1cc79601f3b00ab70d3d0a17db89cea286f84c"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"a7ec499fa2984d6b9f2c116cd0ee6128a352566b47d42486cde3d12b97a0a9c0"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.13088/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations","authors_text":"Fei Ding, Yongkang Zhang, youwei wang, Zijian Zeng","cross_cats":["cs.AI"],"headline":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift.","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-04T02:47:28Z","title":"Design Conditions for Intra-Group Learning of Sequence-Level Rewards: Token Gradient Cancellation"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.13088","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-13T18:38:00.557651Z","id":"b38de77e-6f47-426b-9868-4fb6a5965e2a","model_set":{"reader":"grok-4.3"},"one_line_summary":"Intra-group objectives in sparse-reward RL must maintain token gradient exchangeability to enable cancellation on weak-credit tokens and stabilize training.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Intra-group objectives for sequence rewards must preserve gradient exchangeability across tokens to enable cancellation on weak-credit high-frequency tokens and block reward-irrelevant drift.","strongest_claim":"to prevent reward-irrelevant drift, intra-group objectives must maintain gradient exchangeability across token updates, enabling gradient cancellation on weak-credit/high-frequency tokens.","weakest_assumption":"That the observed training failures (learning tax, solution probability drift, entropy collapse) are primarily caused by disruption of token-level gradient exchangeability rather than other factors such as reward sparsity itself or optimizer dynamics."}},"verdict_id":"b38de77e-6f47-426b-9868-4fb6a5965e2a"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:dda5fcecc740f25bf544e3d8b75e8b3212e2b41b6bc2b6abc011fa2f06ddeaad","target":"record","created_at":"2026-05-26T01:02:34Z","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":"7a24e707f83e0a91d08fdb7046463ad76317ecd108b7c3bcabac36b363346fff","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-04T02:47:28Z","title_canon_sha256":"46d7e74857a28d4a18ccaf7d2c9667372fb0b4eaa2388ddd2b52c1205e6d697b"},"schema_version":"1.0","source":{"id":"2604.13088","kind":"arxiv","version":2}},"canonical_sha256":"8dd8ebe82a30d64f41a07cb8a1a532380dc2694bcd29f380244ef81f424422b5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8dd8ebe82a30d64f41a07cb8a1a532380dc2694bcd29f380244ef81f424422b5","first_computed_at":"2026-05-26T01:02:34.140627Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T01:02:34.140627Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"3WbpDGPf8koi5z0Ph1GP3X8R+fLZODojXrqimzWbLdz3CdG+wRv1sssa5e4upjbPBvO60X5QmxRsMoZZnhtxDw==","signature_status":"signed_v1","signed_at":"2026-05-26T01:02:34.141505Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.13088","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:dda5fcecc740f25bf544e3d8b75e8b3212e2b41b6bc2b6abc011fa2f06ddeaad","sha256:beddecf6754261519eccb5fdaff0cebf6a9c4120f77569a7dd370a337a01bbbd"],"state_sha256":"9e21b401de468714c7de2b90e7a22c5fd14259202b336fa710e719e8f5a6b3d6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"0akX1Nfp16qgBdVSdLjEfxIttDcHwAs6WS2+01YSHxNMytJIxBkLJ8xEHa4fgFDh2V5yUQy930R9YQcDxOtLBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-10T04:49:07.578644Z","bundle_sha256":"d45d30b435adb37dc3349d7b63c4fb210e7b3017b4c0b2ea0573a25051d5b7c4"}}