{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WDZOQVVNWAQLZEAO2JTLSSXCN2","short_pith_number":"pith:WDZOQVVN","schema_version":"1.0","canonical_sha256":"b0f2e856adb020bc900ed266b94ae26ebf90d9262e5e731a1b956b787ac4fa2c","source":{"kind":"arxiv","id":"2501.07888","version":3},"attestation_state":"computed","paper":{"title":"Tarsier2: Advancing Large Vision-Language Models from Detailed Video Description to Comprehensive Video Understanding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Haomiao Sun, Jiawei Wang, Liping Yuan, Yuan Lin, Yuchen Zhang","submitted_at":"2025-01-14T06:54:39Z","abstract_excerpt":"We introduce Tarsier2, a state-of-the-art large vision-language model (LVLM) designed for generating detailed and accurate video descriptions, while also exhibiting superior general video understanding capabilities. Tarsier2 achieves significant advancements through three key upgrades: (1) Scaling pre-training data from 11M to 40M video-text pairs, enriching both volume and diversity; (2) Performing fine-grained temporal alignment during supervised fine-tuning; (3) Using model-based sampling to automatically construct preference data and applying DPO training for optimization. Extensive experi"},"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":"2501.07888","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-01-14T06:54:39Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0a148cb8f76b6b263a06b3e34bf354f17e16a6a705eb5f354943d3d0e03c6647","abstract_canon_sha256":"294c1d0933a2666993702296e6e19004e5c3076b8909927d641c22934ba7022e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:04:41.818541Z","signature_b64":"gOnIzSxWljiAE1TT1r7LHoNOOnnqK2lHcHsLheOYhQLiQILsedOsZH4Mqz2wZqgCEoMGBEtYEWITOeomaBzQAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b0f2e856adb020bc900ed266b94ae26ebf90d9262e5e731a1b956b787ac4fa2c","last_reissued_at":"2026-07-05T10:04:41.818083Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:04:41.818083Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tarsier2: Advancing Large Vision-Language Models from Detailed Video Description to Comprehensive Video Understanding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Haomiao Sun, Jiawei Wang, Liping Yuan, Yuan Lin, Yuchen Zhang","submitted_at":"2025-01-14T06:54:39Z","abstract_excerpt":"We introduce Tarsier2, a state-of-the-art large vision-language model (LVLM) designed for generating detailed and accurate video descriptions, while also exhibiting superior general video understanding capabilities. Tarsier2 achieves significant advancements through three key upgrades: (1) Scaling pre-training data from 11M to 40M video-text pairs, enriching both volume and diversity; (2) Performing fine-grained temporal alignment during supervised fine-tuning; (3) Using model-based sampling to automatically construct preference data and applying DPO training for optimization. Extensive experi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.07888","kind":"arxiv","version":3},"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/2501.07888/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":"2501.07888","created_at":"2026-07-05T10:04:41.818134+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.07888v3","created_at":"2026-07-05T10:04:41.818134+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.07888","created_at":"2026-07-05T10:04:41.818134+00:00"},{"alias_kind":"pith_short_12","alias_value":"WDZOQVVNWAQL","created_at":"2026-07-05T10:04:41.818134+00:00"},{"alias_kind":"pith_short_16","alias_value":"WDZOQVVNWAQLZEAO","created_at":"2026-07-05T10:04:41.818134+00:00"},{"alias_kind":"pith_short_8","alias_value":"WDZOQVVN","created_at":"2026-07-05T10:04:41.818134+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.24636","citing_title":"CineCap: Structured Reasoning with Spatio-Temporal Anchors for Cinematographic Video Captioning","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01667","citing_title":"Temporal and Cross-Modal Alignment for Enhanced Audiovisual Video Captioning","ref_index":66,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09393","citing_title":"CapRL++: Unified Reinforcement Learning with Verifiable Rewards for Dense Image and Video Captioning","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07433","citing_title":"Watch, Remember, Reason: Human-View Video Understanding with MLLMs","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01117","citing_title":"MoHallBench: A Benchmark for Motion Hallucination in Video Large Language Models","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29531","citing_title":"MotionAtlas: Detailed Region Captioning for Motion-Centric Videos","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25979","citing_title":"LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28023","citing_title":"VCap: Hypergeometric Rewards for Weak-to-Strong Visual Captioning","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31075","citing_title":"Task-Focused Memorization for Multimodal Agents","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22962","citing_title":"GazeBehavior Annotation Toolkit (GBAT): AI-powered toolkit for automatic annotation of egocentric eye-tracking and video data of child-caregiver interaction","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09904","citing_title":"TOC-Bench: A Temporal Object Consistency Benchmark for Video Large Language Models","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09904","citing_title":"TOC-Bench: A Temporal Object Consistency Benchmark for Video Large Language Models","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05207","citing_title":"Syn4D: A Multiview Synthetic 4D Dataset","ref_index":130,"is_internal_anchor":false},{"citing_arxiv_id":"2506.09113","citing_title":"Seedance 1.0: Exploring the Boundaries of Video Generation Models","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2506.09985","citing_title":"V-JEPA 2: Self-Supervised Video Models Enable Understanding, Prediction and Planning","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2","json":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2.json","graph_json":"https://pith.science/api/pith-number/WDZOQVVNWAQLZEAO2JTLSSXCN2/graph.json","events_json":"https://pith.science/api/pith-number/WDZOQVVNWAQLZEAO2JTLSSXCN2/events.json","paper":"https://pith.science/paper/WDZOQVVN"},"agent_actions":{"view_html":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2","download_json":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2.json","view_paper":"https://pith.science/paper/WDZOQVVN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.07888&json=true","fetch_graph":"https://pith.science/api/pith-number/WDZOQVVNWAQLZEAO2JTLSSXCN2/graph.json","fetch_events":"https://pith.science/api/pith-number/WDZOQVVNWAQLZEAO2JTLSSXCN2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2/action/storage_attestation","attest_author":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2/action/author_attestation","sign_citation":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2/action/citation_signature","submit_replication":"https://pith.science/pith/WDZOQVVNWAQLZEAO2JTLSSXCN2/action/replication_record"}},"created_at":"2026-07-05T10:04:41.818134+00:00","updated_at":"2026-07-05T10:04:41.818134+00:00"}