{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WS35JDDHSCOBB3CAU4BX7ZT6RY","short_pith_number":"pith:WS35JDDH","schema_version":"1.0","canonical_sha256":"b4b7d48c67909c10ec40a7037fe67e8e112d565e6602238faf6238cd4554fbcf","source":{"kind":"arxiv","id":"2504.02259","version":3},"attestation_state":"computed","paper":{"title":"T*: Re-thinking Temporal Search for Long-Form Video Understanding","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cristobal Eyzaguirre, Ehsan Adeli, Haosen Sun, Jiajun Wu, Jinhui Ye, Juan Carlos Niebles, Keshigeyan Chandrasegaran, Li Fei-Fei, Manling Li, Yonatan Bisk, Zane Durante, Zihan Wang","submitted_at":"2025-04-03T04:03:10Z","abstract_excerpt":"Efficiently understanding long-form videos remains a significant challenge in computer vision. In this work, we revisit temporal search paradigms for long-form video understanding and address a fundamental issue pertaining to all state-of-the-art (SOTA) long-context vision-language models (VLMs). Our contributions are twofold: First, we frame temporal search as a Long Video Haystack problem: finding a minimal set of relevant frames (e.g., one to five) from tens of thousands based on specific queries. Upon this formulation, we introduce LV-Haystack, the first dataset with 480 hours of videos, 1"},"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":"2504.02259","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-04-03T04:03:10Z","cross_cats_sorted":[],"title_canon_sha256":"b4c0a230889779716cf6f8555dd3e77fde042424ec42eb1863a2760713a83b08","abstract_canon_sha256":"cdd69051996b0d73f4cd81bfdb2683df570521c1a8f85bd95e5fd636d2c1bade"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:58:26.965478Z","signature_b64":"WUDZuO7Bi3/kQMKPrd2GMgKv6qstqy6o50jtATpVgP6g3xAAwmJXw03HeWGgNjWXJAOcjiGxYiiRjSzA/JwsDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b4b7d48c67909c10ec40a7037fe67e8e112d565e6602238faf6238cd4554fbcf","last_reissued_at":"2026-07-05T11:58:26.964969Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:58:26.964969Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"T*: Re-thinking Temporal Search for Long-Form Video Understanding","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Cristobal Eyzaguirre, Ehsan Adeli, Haosen Sun, Jiajun Wu, Jinhui Ye, Juan Carlos Niebles, Keshigeyan Chandrasegaran, Li Fei-Fei, Manling Li, Yonatan Bisk, Zane Durante, Zihan Wang","submitted_at":"2025-04-03T04:03:10Z","abstract_excerpt":"Efficiently understanding long-form videos remains a significant challenge in computer vision. In this work, we revisit temporal search paradigms for long-form video understanding and address a fundamental issue pertaining to all state-of-the-art (SOTA) long-context vision-language models (VLMs). Our contributions are twofold: First, we frame temporal search as a Long Video Haystack problem: finding a minimal set of relevant frames (e.g., one to five) from tens of thousands based on specific queries. Upon this formulation, we introduce LV-Haystack, the first dataset with 480 hours of videos, 1"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.02259","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/2504.02259/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":"2504.02259","created_at":"2026-07-05T11:58:26.965028+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.02259v3","created_at":"2026-07-05T11:58:26.965028+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.02259","created_at":"2026-07-05T11:58:26.965028+00:00"},{"alias_kind":"pith_short_12","alias_value":"WS35JDDHSCOB","created_at":"2026-07-05T11:58:26.965028+00:00"},{"alias_kind":"pith_short_16","alias_value":"WS35JDDHSCOBB3CA","created_at":"2026-07-05T11:58:26.965028+00:00"},{"alias_kind":"pith_short_8","alias_value":"WS35JDDH","created_at":"2026-07-05T11:58:26.965028+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20515","citing_title":"S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09223","citing_title":"CREST: Curvature-Regulated Event-Centric Sampling for Efficient Long-Video Understanding","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20515","citing_title":"S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31598","citing_title":"Linear Scaling Video VLMs for Long Video Understanding","ref_index":75,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09223","citing_title":"CREST: Curvature-Regulated Event-Centric Sampling for Efficient Long-Video Understanding","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY","json":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY.json","graph_json":"https://pith.science/api/pith-number/WS35JDDHSCOBB3CAU4BX7ZT6RY/graph.json","events_json":"https://pith.science/api/pith-number/WS35JDDHSCOBB3CAU4BX7ZT6RY/events.json","paper":"https://pith.science/paper/WS35JDDH"},"agent_actions":{"view_html":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY","download_json":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY.json","view_paper":"https://pith.science/paper/WS35JDDH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.02259&json=true","fetch_graph":"https://pith.science/api/pith-number/WS35JDDHSCOBB3CAU4BX7ZT6RY/graph.json","fetch_events":"https://pith.science/api/pith-number/WS35JDDHSCOBB3CAU4BX7ZT6RY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY/action/storage_attestation","attest_author":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY/action/author_attestation","sign_citation":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY/action/citation_signature","submit_replication":"https://pith.science/pith/WS35JDDHSCOBB3CAU4BX7ZT6RY/action/replication_record"}},"created_at":"2026-07-05T11:58:26.965028+00:00","updated_at":"2026-07-05T11:58:26.965028+00:00"}