{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HBQNWEDYQHMKTIPES7MNUPYUEU","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":"65fbf67244486fa2246b4d2fada7fcf3f61708833de83bfe592dc929f3070947","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:55:18Z","title_canon_sha256":"3e63b5f0621ec24924545486c7cccd4d1e39f027abe3486fe1716415d604e51d"},"schema_version":"1.0","source":{"id":"2605.13467","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13467","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13467v1","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13467","created_at":"2026-05-18T02:44:41Z"},{"alias_kind":"pith_short_12","alias_value":"HBQNWEDYQHMK","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HBQNWEDYQHMKTIPE","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HBQNWEDY","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:841b83489be26417bbe088903f1b7c0b13c83150023f7465ee69873b99affea1","target":"graph","created_at":"2026-05-18T02:44:41Z","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":"We demonstrate that PDCR outperforms the naive, global-reward formulation and sparse-reward baselines on key V-L reasoning benchmarks."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The unsupervised clustering driven by the model-internal Visual Dependence Score accurately partitions steps into perception and reasoning without any labeled supervision or external verification."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"PDCR improves vision-language reasoning by computing separate normalized confidence advantages for perception steps and reasoning steps after unsupervised decomposition."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Decomposing confidence rewards into perception and reasoning clusters improves vision-language model training."}],"snapshot_sha256":"714cd5196fa51bd7896d97721fa492f8bbceff8d1586e8100f653db7bb716ee2"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Reinforcement Learning with Verifiable Rewards (RLVR) traditionally relies on a sparse, outcome-based signal. Recent work shows that providing a fine-grained, model-intrinsic signal (rewarding the confidence growth in the ground-truth answer) effectively improves language reasoning training by providing step-level guidance without costly external models. While effective for unimodal text, we find that naively applying this global reward to vision-language (V-L) reasoning is a suboptimal strategy, as the task is a heterogeneous mix of sparse visual perception and dense textual reasoning. This g","authors_text":"Chang D. Yoo, Chong Luo, Eunseop Yoon, Gwanhyeong Koo, Hee Suk Yoon, Ji Woo Hong, Mark Hasegawa-Johnson, Qi Dai, SooHwan Eom","cross_cats":[],"headline":"Decomposing confidence rewards into perception and reasoning clusters improves vision-language model training.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:55:18Z","title":"PDCR: Perception-Decomposed Confidence Reward for Vision-Language Reasoning"},"references":{"count":79,"internal_anchors":12,"resolved_work":79,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Gpt-4v(ision) system card. 2023. 1","work_id":"a549d3c1-1572-4949-8f9e-de9af53432a9","year":2023},{"cited_arxiv_id":"2502.13923","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Vrprm: Process reward modeling via visual reasoning.arXiv preprint arXiv:2508.03556, 2025","work_id":"673ea4d5-7e09-4f7f-acf3-106a8e69f259","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"arXiv preprint arXiv:2505.23558 , year=","work_id":"cc9e5e98-c226-4e12-bf74-ff438225c4c0","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Ultrafeedback: Boosting language models with scaled ai feedback","work_id":"bc448bac-e570-4911-b8b6-fc2bf33a532a","year":null}],"snapshot_sha256":"eaa94caf858d0e6c7ba30d5f326c505e9c792d75883c64c48a8460a16dd29639"},"source":{"id":"2605.13467","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:16:10.963422Z","id":"d9383de8-b038-4d3c-bb25-311e4729b4d7","model_set":{"reader":"grok-4.3"},"one_line_summary":"PDCR improves vision-language reasoning by computing separate normalized confidence advantages for perception steps and reasoning steps after unsupervised decomposition.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Decomposing confidence rewards into perception and reasoning clusters improves vision-language model training.","strongest_claim":"We demonstrate that PDCR outperforms the naive, global-reward formulation and sparse-reward baselines on key V-L reasoning benchmarks.","weakest_assumption":"The unsupervised clustering driven by the model-internal Visual Dependence Score accurately partitions steps into perception and reasoning without any labeled supervision or external verification."}},"verdict_id":"d9383de8-b038-4d3c-bb25-311e4729b4d7"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:beac8b2b221403aec0cd8b6d5389a7fdde88167af744b5db9d61e7cd6b28736e","target":"record","created_at":"2026-05-18T02:44:41Z","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":"65fbf67244486fa2246b4d2fada7fcf3f61708833de83bfe592dc929f3070947","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:55:18Z","title_canon_sha256":"3e63b5f0621ec24924545486c7cccd4d1e39f027abe3486fe1716415d604e51d"},"schema_version":"1.0","source":{"id":"2605.13467","kind":"arxiv","version":1}},"canonical_sha256":"3860db107881d8a9a1e497d8da3f142520076fb5acc35d29893e75cd61db867b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3860db107881d8a9a1e497d8da3f142520076fb5acc35d29893e75cd61db867b","first_computed_at":"2026-05-18T02:44:41.611045Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:44:41.611045Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"XGS+T65y99iAewShMz7PQwQIyQiDxRjlNDrYrOMNkPRTpf7OUaB2qio2adZshUonn2Ku7jqVZn+yoMicYeYiDg==","signature_status":"signed_v1","signed_at":"2026-05-18T02:44:41.611661Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13467","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:beac8b2b221403aec0cd8b6d5389a7fdde88167af744b5db9d61e7cd6b28736e","sha256:841b83489be26417bbe088903f1b7c0b13c83150023f7465ee69873b99affea1"],"state_sha256":"34f75b4d6a0c337222d69daa9bf68e52bda6ac521ac95f48e97df5e9a8f5e1c3"}