{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:R7JSIKZ5HNYMVEEYVMGRXFAIXF","short_pith_number":"pith:R7JSIKZ5","schema_version":"1.0","canonical_sha256":"8fd3242b3d3b70ca9098ab0d1b9408b974a605e5eb28e1dc72f09b445dc344f2","source":{"kind":"arxiv","id":"2404.04346","version":3},"attestation_state":"computed","paper":{"title":"Koala: Key frame-conditioned long video-LLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bryan A. Plummer, Bryan Russell, Hanieh Deilamsalehy, Jui-Hsien Wang, Kate Saenko, Ping Hu, Reuben Tan, Ximeng Sun","submitted_at":"2024-04-05T18:33:04Z","abstract_excerpt":"Long video question answering is a challenging task that involves recognizing short-term activities and reasoning about their fine-grained relationships. State-of-the-art video Large Language Models (vLLMs) hold promise as a viable solution due to their demonstrated emergent capabilities on new tasks. However, despite being trained on millions of short seconds-long videos, vLLMs are unable to understand minutes-long videos and accurately answer questions about them. To address this limitation, we propose a lightweight and self-supervised approach, Key frame-conditioned long video-LLM (Koala), "},"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":"2404.04346","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-04-05T18:33:04Z","cross_cats_sorted":[],"title_canon_sha256":"b652313124559341886d34e817d822733a2570ad7b710e8af9688125d6afdc6d","abstract_canon_sha256":"846321f956fc09ecde19c21359fe97233dd8f6fb3917219a9c4ef77cdfdcad58"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:15:12.778297Z","signature_b64":"LOFNq73A/4fBfyXlVr4PRAjSZjPgv8bfiO718P66cxg/voZMXmEt0VI1ocVlvcToFkHnC3Y1X6VJuIHeLskMBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8fd3242b3d3b70ca9098ab0d1b9408b974a605e5eb28e1dc72f09b445dc344f2","last_reissued_at":"2026-07-05T08:15:12.777835Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:15:12.777835Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Koala: Key frame-conditioned long video-LLM","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bryan A. Plummer, Bryan Russell, Hanieh Deilamsalehy, Jui-Hsien Wang, Kate Saenko, Ping Hu, Reuben Tan, Ximeng Sun","submitted_at":"2024-04-05T18:33:04Z","abstract_excerpt":"Long video question answering is a challenging task that involves recognizing short-term activities and reasoning about their fine-grained relationships. State-of-the-art video Large Language Models (vLLMs) hold promise as a viable solution due to their demonstrated emergent capabilities on new tasks. However, despite being trained on millions of short seconds-long videos, vLLMs are unable to understand minutes-long videos and accurately answer questions about them. To address this limitation, we propose a lightweight and self-supervised approach, Key frame-conditioned long video-LLM (Koala), "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.04346","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/2404.04346/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":"2404.04346","created_at":"2026-07-05T08:15:12.777890+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.04346v3","created_at":"2026-07-05T08:15:12.777890+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.04346","created_at":"2026-07-05T08:15:12.777890+00:00"},{"alias_kind":"pith_short_12","alias_value":"R7JSIKZ5HNYM","created_at":"2026-07-05T08:15:12.777890+00:00"},{"alias_kind":"pith_short_16","alias_value":"R7JSIKZ5HNYMVEEY","created_at":"2026-07-05T08:15:12.777890+00:00"},{"alias_kind":"pith_short_8","alias_value":"R7JSIKZ5","created_at":"2026-07-05T08:15:12.777890+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21734","citing_title":"HPP: Hierarchical Programmatic Probing for Long Video Understanding by Decoupling Perception and Reasoning","ref_index":169,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF","json":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF.json","graph_json":"https://pith.science/api/pith-number/R7JSIKZ5HNYMVEEYVMGRXFAIXF/graph.json","events_json":"https://pith.science/api/pith-number/R7JSIKZ5HNYMVEEYVMGRXFAIXF/events.json","paper":"https://pith.science/paper/R7JSIKZ5"},"agent_actions":{"view_html":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF","download_json":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF.json","view_paper":"https://pith.science/paper/R7JSIKZ5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.04346&json=true","fetch_graph":"https://pith.science/api/pith-number/R7JSIKZ5HNYMVEEYVMGRXFAIXF/graph.json","fetch_events":"https://pith.science/api/pith-number/R7JSIKZ5HNYMVEEYVMGRXFAIXF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF/action/storage_attestation","attest_author":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF/action/author_attestation","sign_citation":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF/action/citation_signature","submit_replication":"https://pith.science/pith/R7JSIKZ5HNYMVEEYVMGRXFAIXF/action/replication_record"}},"created_at":"2026-07-05T08:15:12.777890+00:00","updated_at":"2026-07-05T08:15:12.777890+00:00"}