{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YVOGZY6Q7JZI5S5OTF2SXXYLEA","short_pith_number":"pith:YVOGZY6Q","schema_version":"1.0","canonical_sha256":"c55c6ce3d0fa728ecbae99752bdf0b20391cc1ace62c55d4b2150a8d1bdc55ef","source":{"kind":"arxiv","id":"2505.19702","version":1},"attestation_state":"computed","paper":{"title":"Point-RFT: Improving Multimodal Reasoning with Visually Grounded Reinforcement Finetuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chung-Ching Lin, Kevin Lin, Lijuan Wang, Linjie Li, Minheng Ni, Wangmeng Zuo, Zhengyuan Yang","submitted_at":"2025-05-26T08:54:14Z","abstract_excerpt":"Recent advances in large language models have significantly improved textual reasoning through the effective use of Chain-of-Thought (CoT) and reinforcement learning. However, extending these successes to vision-language tasks remains challenging due to inherent limitations in text-only CoT, such as visual hallucinations and insufficient multimodal integration. In this paper, we introduce Point-RFT, a multimodal reasoning framework explicitly designed to leverage visually grounded CoT reasoning for visual document understanding. Our approach consists of two stages: First, we conduct format fin"},"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":"2505.19702","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-05-26T08:54:14Z","cross_cats_sorted":[],"title_canon_sha256":"0200e2a032305331fc3657c3be6bb932dade8d4030a34d71bfc043693a0b56f3","abstract_canon_sha256":"35f5c8d510cec0afbad0c79f4bd80ce0c66e6209a6af2a95800cb929c3e953e4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:09:34.780829Z","signature_b64":"FzcL5JOfK9EqXZHZsjzzj9DWDIMk1j90LCEtkEEvU2nOHShzujwh/3l/q++T5Ui2eDgE4UvJ0d4ImoIPAquDBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c55c6ce3d0fa728ecbae99752bdf0b20391cc1ace62c55d4b2150a8d1bdc55ef","last_reissued_at":"2026-07-05T11:09:34.780330Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:09:34.780330Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Point-RFT: Improving Multimodal Reasoning with Visually Grounded Reinforcement Finetuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chung-Ching Lin, Kevin Lin, Lijuan Wang, Linjie Li, Minheng Ni, Wangmeng Zuo, Zhengyuan Yang","submitted_at":"2025-05-26T08:54:14Z","abstract_excerpt":"Recent advances in large language models have significantly improved textual reasoning through the effective use of Chain-of-Thought (CoT) and reinforcement learning. However, extending these successes to vision-language tasks remains challenging due to inherent limitations in text-only CoT, such as visual hallucinations and insufficient multimodal integration. In this paper, we introduce Point-RFT, a multimodal reasoning framework explicitly designed to leverage visually grounded CoT reasoning for visual document understanding. Our approach consists of two stages: First, we conduct format fin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.19702","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/2505.19702/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":"2505.19702","created_at":"2026-07-05T11:09:34.780390+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.19702v1","created_at":"2026-07-05T11:09:34.780390+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.19702","created_at":"2026-07-05T11:09:34.780390+00:00"},{"alias_kind":"pith_short_12","alias_value":"YVOGZY6Q7JZI","created_at":"2026-07-05T11:09:34.780390+00:00"},{"alias_kind":"pith_short_16","alias_value":"YVOGZY6Q7JZI5S5O","created_at":"2026-07-05T11:09:34.780390+00:00"},{"alias_kind":"pith_short_8","alias_value":"YVOGZY6Q","created_at":"2026-07-05T11:09:34.780390+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.22192","citing_title":"CharTide: Data-Centric Chart-to-Code Generation via Tri-Perspective Tuning and Inquiry-Driven Evolution","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19927","citing_title":"CARE: Competence-Aware Reward Shaping for Adaptive Reasoning Length in Video-MLLMs","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24251","citing_title":"Latent Visual Reasoning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03307","citing_title":"V-Reflection: Transforming MLLMs from Passive Observers to Active Interrogators","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11774","citing_title":"From Token to Token Pair: Efficient Prompt Compression for Large Language Models in Clinical Prediction","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22192","citing_title":"CharTide: Data-Centric Chart-to-Code Generation via Tri-Perspective Tuning and Inquiry-Driven Evolution","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA","json":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA.json","graph_json":"https://pith.science/api/pith-number/YVOGZY6Q7JZI5S5OTF2SXXYLEA/graph.json","events_json":"https://pith.science/api/pith-number/YVOGZY6Q7JZI5S5OTF2SXXYLEA/events.json","paper":"https://pith.science/paper/YVOGZY6Q"},"agent_actions":{"view_html":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA","download_json":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA.json","view_paper":"https://pith.science/paper/YVOGZY6Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.19702&json=true","fetch_graph":"https://pith.science/api/pith-number/YVOGZY6Q7JZI5S5OTF2SXXYLEA/graph.json","fetch_events":"https://pith.science/api/pith-number/YVOGZY6Q7JZI5S5OTF2SXXYLEA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA/action/storage_attestation","attest_author":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA/action/author_attestation","sign_citation":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA/action/citation_signature","submit_replication":"https://pith.science/pith/YVOGZY6Q7JZI5S5OTF2SXXYLEA/action/replication_record"}},"created_at":"2026-07-05T11:09:34.780390+00:00","updated_at":"2026-07-05T11:09:34.780390+00:00"}