{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2KZDX5K4XMCGZ3WIM36VMVWLTK","short_pith_number":"pith:2KZDX5K4","schema_version":"1.0","canonical_sha256":"d2b23bf55cbb046ceec866fd5656cb9a85b68d676d8291fef20c7da5691030b7","source":{"kind":"arxiv","id":"2503.07478","version":1},"attestation_state":"computed","paper":{"title":"VLRMBench: A Comprehensive and Challenging Benchmark for Vision-Language Reward Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Daoxin Zhang, Jiacheng Ruan, Ting Liu, Wenzhen Yuan, Xian Gao, Yao Hu, Ye Guo, Yuzhuo Fu, Zhe Xu","submitted_at":"2025-03-10T15:52:57Z","abstract_excerpt":"Although large visual-language models (LVLMs) have demonstrated strong performance in multimodal tasks, errors may occasionally arise due to biases during the reasoning process. Recently, reward models (RMs) have become increasingly pivotal in the reasoning process. Specifically, process RMs evaluate each reasoning step, outcome RMs focus on the assessment of reasoning results, and critique RMs perform error analysis on the entire reasoning process, followed by corrections. However, existing benchmarks for vision-language RMs (VLRMs) typically assess only a single aspect of their capabilities "},"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":false},"canonical_record":{"source":{"id":"2503.07478","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-10T15:52:57Z","cross_cats_sorted":[],"title_canon_sha256":"4f2aec3c17e61e0309291a66b6027c4b8d9632574f98c828b6e2766f3be2de86","abstract_canon_sha256":"5030d569d60b053ce438db2c2b90e008602dfaf89cfa7c7552588246f0d1bd74"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:27:57.131389Z","signature_b64":"9UVO2FhvD0tBSsegx/HHbBGK4FdZkcGmsXdObZBcr7GyjHotUWiTwhhfHmS4cuVv1aipFVlNRNEr+mm0jHVfAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d2b23bf55cbb046ceec866fd5656cb9a85b68d676d8291fef20c7da5691030b7","last_reissued_at":"2026-07-05T10:27:57.130813Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:27:57.130813Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VLRMBench: A Comprehensive and Challenging Benchmark for Vision-Language Reward Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Daoxin Zhang, Jiacheng Ruan, Ting Liu, Wenzhen Yuan, Xian Gao, Yao Hu, Ye Guo, Yuzhuo Fu, Zhe Xu","submitted_at":"2025-03-10T15:52:57Z","abstract_excerpt":"Although large visual-language models (LVLMs) have demonstrated strong performance in multimodal tasks, errors may occasionally arise due to biases during the reasoning process. Recently, reward models (RMs) have become increasingly pivotal in the reasoning process. Specifically, process RMs evaluate each reasoning step, outcome RMs focus on the assessment of reasoning results, and critique RMs perform error analysis on the entire reasoning process, followed by corrections. However, existing benchmarks for vision-language RMs (VLRMs) typically assess only a single aspect of their capabilities "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.07478","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/2503.07478/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2503.07478","created_at":"2026-07-05T10:27:57.130872+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.07478v1","created_at":"2026-07-05T10:27:57.130872+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.07478","created_at":"2026-07-05T10:27:57.130872+00:00"},{"alias_kind":"pith_short_12","alias_value":"2KZDX5K4XMCG","created_at":"2026-07-05T10:27:57.130872+00:00"},{"alias_kind":"pith_short_16","alias_value":"2KZDX5K4XMCGZ3WI","created_at":"2026-07-05T10:27:57.130872+00:00"},{"alias_kind":"pith_short_8","alias_value":"2KZDX5K4","created_at":"2026-07-05T10:27:57.130872+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2504.12501","citing_title":"Reinforcement Learning from Human Feedback","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2506.01937","citing_title":"RewardBench 2: Advancing Reward Model Evaluation","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05445","citing_title":"Learning What Matters: Dynamic Dimension Selection and Aggregation for Interpretable Vision-Language Reward Modeling","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK","json":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK.json","graph_json":"https://pith.science/api/pith-number/2KZDX5K4XMCGZ3WIM36VMVWLTK/graph.json","events_json":"https://pith.science/api/pith-number/2KZDX5K4XMCGZ3WIM36VMVWLTK/events.json","paper":"https://pith.science/paper/2KZDX5K4"},"agent_actions":{"view_html":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK","download_json":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK.json","view_paper":"https://pith.science/paper/2KZDX5K4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.07478&json=true","fetch_graph":"https://pith.science/api/pith-number/2KZDX5K4XMCGZ3WIM36VMVWLTK/graph.json","fetch_events":"https://pith.science/api/pith-number/2KZDX5K4XMCGZ3WIM36VMVWLTK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK/action/storage_attestation","attest_author":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK/action/author_attestation","sign_citation":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK/action/citation_signature","submit_replication":"https://pith.science/pith/2KZDX5K4XMCGZ3WIM36VMVWLTK/action/replication_record"}},"created_at":"2026-07-05T10:27:57.130872+00:00","updated_at":"2026-07-05T10:27:57.130872+00:00"}