{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SQKJEUHT7NUGIUIGX7CO6HVTOG","short_pith_number":"pith:SQKJEUHT","schema_version":"1.0","canonical_sha256":"94149250f3fb68645106bfc4ef1eb3719b0a7efbe9b3bc2df7b5cc1acad4ebe5","source":{"kind":"arxiv","id":"2503.18013","version":1},"attestation_state":"computed","paper":{"title":"Vision-R1: Evolving Human-Free Alignment in Large Vision-Language Models via Vision-Guided Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Fan Yang, Hongyin Zhao, Jinqiao Wang, Ming Tang, Shurong Zheng, Yousong Zhu, Yufei Zhan","submitted_at":"2025-03-23T10:21:14Z","abstract_excerpt":"Large Vision-Language Models (LVLMs) typically follow a two-stage training paradigm-pretraining and supervised fine-tuning. Recently, preference optimization, derived from the language domain, has emerged as an effective post-training reinforcement strategy to enhance capabilities of LVLMs. However, constructing high-quality human-annotated preference data and developing robust reward models to mimic these preferences are both costly and challenging. Motivated by this observation, we propose Vision-R1, a novel vision-guided R1-like reinforcement learning algorithm for LVLMs that rewards models"},"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":"2503.18013","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-23T10:21:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"fde22600d754d7905e49e34ef3e6293f30d7ac60127819cafad6e7e4b05a85ca","abstract_canon_sha256":"d99ceb87c4d3374785b5626180647d8b8ce17771f7fe9b3ad18d7273cbc853c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:38:00.442871Z","signature_b64":"fxbbn7K0GNIUh05gZ7luwxcfKvntDvmZyAa2esOsgdJkheZ+eCEka6jEu2nXBOVjKjEMtZvO73nS/ITyA2jkBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"94149250f3fb68645106bfc4ef1eb3719b0a7efbe9b3bc2df7b5cc1acad4ebe5","last_reissued_at":"2026-07-05T10:38:00.442372Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:38:00.442372Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Vision-R1: Evolving Human-Free Alignment in Large Vision-Language Models via Vision-Guided Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Fan Yang, Hongyin Zhao, Jinqiao Wang, Ming Tang, Shurong Zheng, Yousong Zhu, Yufei Zhan","submitted_at":"2025-03-23T10:21:14Z","abstract_excerpt":"Large Vision-Language Models (LVLMs) typically follow a two-stage training paradigm-pretraining and supervised fine-tuning. Recently, preference optimization, derived from the language domain, has emerged as an effective post-training reinforcement strategy to enhance capabilities of LVLMs. However, constructing high-quality human-annotated preference data and developing robust reward models to mimic these preferences are both costly and challenging. Motivated by this observation, we propose Vision-R1, a novel vision-guided R1-like reinforcement learning algorithm for LVLMs that rewards models"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.18013","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/2503.18013/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":"2503.18013","created_at":"2026-07-05T10:38:00.442433+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.18013v1","created_at":"2026-07-05T10:38:00.442433+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.18013","created_at":"2026-07-05T10:38:00.442433+00:00"},{"alias_kind":"pith_short_12","alias_value":"SQKJEUHT7NUG","created_at":"2026-07-05T10:38:00.442433+00:00"},{"alias_kind":"pith_short_16","alias_value":"SQKJEUHT7NUGIUIG","created_at":"2026-07-05T10:38:00.442433+00:00"},{"alias_kind":"pith_short_8","alias_value":"SQKJEUHT","created_at":"2026-07-05T10:38:00.442433+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26196","citing_title":"From Structure to Synergy: A Survey of Vision-Language Perception Paradigm Evolution in Multimodal Large Language Models","ref_index":172,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23557","citing_title":"Dense Reward for Multi-View 3D Reasoning with Global Maps and Local Views","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09711","citing_title":"Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization","ref_index":280,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06294","citing_title":"Towards One-to-Many Temporal Grounding","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2507.00748","citing_title":"Improving the Reasoning of Multi-Image Grounding in MLLMs via Reinforcement Learning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2504.06958","citing_title":"VideoChat-R1: Enhancing Spatio-Temporal Perception via Reinforcement Fine-Tuning","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2602.20913","citing_title":"LongVideo-R1: Smart Navigation for Low-cost Long Video Understanding","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03485","citing_title":"MHPR: Multidimensional Human Perception and Reasoning Benchmark for Large Vision-Languate Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06121","citing_title":"Pest-Thinker: Learning to Think and Reason like Entomologists via Reinforcement Learning","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01324","citing_title":"Beyond Perceptual Shortcuts: Causal-Inspired Debiasing Optimization for Generalizable Video Reasoning in Lightweight MLLMs","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19218","citing_title":"Thinking Before Matching: A Reinforcement Reasoning Paradigm Towards General Person Re-Identification","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08230","citing_title":"Generalization Under Scrutiny: Cross-Domain Detection Progresses, Pitfalls, and Persistent Challenges","ref_index":117,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04379","citing_title":"Reinforce to Learn, Elect to Reason: A Dual Paradigm for Video Reasoning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13602","citing_title":"Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges","ref_index":192,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG","json":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG.json","graph_json":"https://pith.science/api/pith-number/SQKJEUHT7NUGIUIGX7CO6HVTOG/graph.json","events_json":"https://pith.science/api/pith-number/SQKJEUHT7NUGIUIGX7CO6HVTOG/events.json","paper":"https://pith.science/paper/SQKJEUHT"},"agent_actions":{"view_html":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG","download_json":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG.json","view_paper":"https://pith.science/paper/SQKJEUHT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.18013&json=true","fetch_graph":"https://pith.science/api/pith-number/SQKJEUHT7NUGIUIGX7CO6HVTOG/graph.json","fetch_events":"https://pith.science/api/pith-number/SQKJEUHT7NUGIUIGX7CO6HVTOG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG/action/storage_attestation","attest_author":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG/action/author_attestation","sign_citation":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG/action/citation_signature","submit_replication":"https://pith.science/pith/SQKJEUHT7NUGIUIGX7CO6HVTOG/action/replication_record"}},"created_at":"2026-07-05T10:38:00.442433+00:00","updated_at":"2026-07-05T10:38:00.442433+00:00"}