{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:SMIYT6JF3RT33IL33OCLGSRXKV","short_pith_number":"pith:SMIYT6JF","schema_version":"1.0","canonical_sha256":"931189f925dc67bda17bdb84b34a3755589b0d877512ba1d10f86db0cb49d7d3","source":{"kind":"arxiv","id":"2604.11056","version":2},"attestation_state":"computed","paper":{"title":"Where Hindsight Credit Can Reside: A Signed-Capacity View of Token Updates in RLVR","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"The credit a token can carry in RLVR is upper-bounded by its entropy.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hange Zhou, Haodong Wu, Hongyu Ge, Keyi Wu, Qihong Lin, Siyi Liu, Yongqi Zhang, Yuhang He, Zhuo Zheng, Zixin Zhong","submitted_at":"2026-04-13T06:32:49Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) improves the reasoning ability of Large Language Models (LLMs), but sparse outcome rewards make token-level credit assignment difficult.\n  We study token-level credit as a reward-conditioned shift from the behavior policy to a hindsight posterior.\n  In autoregressive RLVR, this shift can be expressed through Conditional Mutual Information (CMI), which shows that token entropy upper-bounds possible hindsight credit.\n  Entropy, however, indicates capacity rather than update direction, so we introduce the Four Quadrant Decomposition to separat"},"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":true},"canonical_record":{"source":{"id":"2604.11056","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-13T06:32:49Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b19507d8506e7f744dae6da2c1140f931113ce7594dd9cc78600704295f35401","abstract_canon_sha256":"ec7e168f367dde92704456449ca2533fa1aa887dd58db9ffcba34e5a2d3477eb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:54.302481Z","signature_b64":"vwfCiZeGKxZHUpTnz1vUeAUsENuIo6eCEsfedTCeV2tu5AEC90PKO7p9KtY4Ks0CyshQsFxBnAA64rJcX1wQCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"931189f925dc67bda17bdb84b34a3755589b0d877512ba1d10f86db0cb49d7d3","last_reissued_at":"2026-05-27T01:05:54.301633Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:54.301633Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Where Hindsight Credit Can Reside: A Signed-Capacity View of Token Updates in RLVR","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"The credit a token can carry in RLVR is upper-bounded by its entropy.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hange Zhou, Haodong Wu, Hongyu Ge, Keyi Wu, Qihong Lin, Siyi Liu, Yongqi Zhang, Yuhang He, Zhuo Zheng, Zixin Zhong","submitted_at":"2026-04-13T06:32:49Z","abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) improves the reasoning ability of Large Language Models (LLMs), but sparse outcome rewards make token-level credit assignment difficult.\n  We study token-level credit as a reward-conditioned shift from the behavior policy to a hindsight posterior.\n  In autoregressive RLVR, this shift can be expressed through Conditional Mutual Information (CMI), which shows that token entropy upper-bounds possible hindsight credit.\n  Entropy, however, indicates capacity rather than update direction, so we introduce the Four Quadrant Decomposition to separat"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We adapt Conditional Mutual Information to the autoregressive RLVR setting and prove that the credit a token can carry is upper-bounded by its entropy.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The adaptation of conditional mutual information to the autoregressive RLVR setting correctly captures the credit a token can carry, and that high-entropy tokens are the primary locus of reasoning improvements.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Token credit in RLVR is upper-bounded by entropy, with reasoning gains concentrated in high-entropy tokens, motivating Entropy-Aware Policy Optimization that outperforms baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"The credit a token can carry in RLVR is upper-bounded by its entropy.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"760faad15b650b36e68df48d88518293dae3236d59b670f2a19f13858e515ac0"},"source":{"id":"2604.11056","kind":"arxiv","version":2},"verdict":{"id":"232c2a53-2d87-48e3-9350-a884ce6b167d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:29:49.448633Z","strongest_claim":"We adapt Conditional Mutual Information to the autoregressive RLVR setting and prove that the credit a token can carry is upper-bounded by its entropy.","one_line_summary":"Token credit in RLVR is upper-bounded by entropy, with reasoning gains concentrated in high-entropy tokens, motivating Entropy-Aware Policy Optimization that outperforms baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The adaptation of conditional mutual information to the autoregressive RLVR setting correctly captures the credit a token can carry, and that high-entropy tokens are the primary locus of reasoning improvements.","pith_extraction_headline":"The credit a token can carry in RLVR is upper-bounded by its entropy."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.11056/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":2,"snapshot_sha256":"ee8768e998565f82a9a18e81fffef09cde0faf3c6ef275744b7e36c13208b9c3"},"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":"2604.11056","created_at":"2026-05-27T01:05:54.301766+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.11056v2","created_at":"2026-05-27T01:05:54.301766+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.11056","created_at":"2026-05-27T01:05:54.301766+00:00"},{"alias_kind":"pith_short_12","alias_value":"SMIYT6JF3RT3","created_at":"2026-05-27T01:05:54.301766+00:00"},{"alias_kind":"pith_short_16","alias_value":"SMIYT6JF3RT33IL3","created_at":"2026-05-27T01:05:54.301766+00:00"},{"alias_kind":"pith_short_8","alias_value":"SMIYT6JF","created_at":"2026-05-27T01:05:54.301766+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2605.22263","citing_title":"Tailoring Teaching to Aptitude: Direction-Adaptive Self-Distillation for LLM Reasoning","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12944","citing_title":"From Instance Selection to Fixed-Pool Data Recipe Search for Supervised Fine-Tuning","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07660","citing_title":"Not All Tokens Learn Alike: Attention Entropy Reveals Heterogeneous Signals in RL Reasoning","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV","json":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV.json","graph_json":"https://pith.science/api/pith-number/SMIYT6JF3RT33IL33OCLGSRXKV/graph.json","events_json":"https://pith.science/api/pith-number/SMIYT6JF3RT33IL33OCLGSRXKV/events.json","paper":"https://pith.science/paper/SMIYT6JF"},"agent_actions":{"view_html":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV","download_json":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV.json","view_paper":"https://pith.science/paper/SMIYT6JF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.11056&json=true","fetch_graph":"https://pith.science/api/pith-number/SMIYT6JF3RT33IL33OCLGSRXKV/graph.json","fetch_events":"https://pith.science/api/pith-number/SMIYT6JF3RT33IL33OCLGSRXKV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV/action/storage_attestation","attest_author":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV/action/author_attestation","sign_citation":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV/action/citation_signature","submit_replication":"https://pith.science/pith/SMIYT6JF3RT33IL33OCLGSRXKV/action/replication_record"}},"created_at":"2026-05-27T01:05:54.301766+00:00","updated_at":"2026-05-27T01:05:54.301766+00:00"}