{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:V7FUB7I2WXMBT4D6LDL2A4COWK","short_pith_number":"pith:V7FUB7I2","schema_version":"1.0","canonical_sha256":"afcb40fd1ab5d819f07e58d7a0704eb2b265ba7720a1d513630625bf7e1a1457","source":{"kind":"arxiv","id":"2503.18943","version":2},"attestation_state":"computed","paper":{"title":"SlowFast-LLaVA-1.5: A Family of Token-Efficient Video Large Language Models for Long-Form Video Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Jiasen Lu, Kai Kang, Meng Cao, Mingfei Gao, Mingze Xu, Shiyu Li, Yinfei Yang, Zhe Gan, Zhengfeng Lai","submitted_at":"2025-03-24T17:59:07Z","abstract_excerpt":"We introduce SlowFast-LLaVA-1.5 (abbreviated as SF-LLaVA-1.5), a family of video large language models (LLMs) offering a token-efficient solution for long-form video understanding. We incorporate the two-stream SlowFast mechanism into a streamlined training pipeline, and perform joint video-image training on a carefully curated data mixture of only publicly available datasets. Our primary focus is on highly efficient model scales (1B and 3B), demonstrating that even relatively small Video LLMs can achieve state-of-the-art performance on video understanding, meeting the demand for mobile-friend"},"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":"2503.18943","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-24T17:59:07Z","cross_cats_sorted":[],"title_canon_sha256":"53256e2a61571d382b75127c03d9eefc905f0704f1356a177225c7d3d38f3f66","abstract_canon_sha256":"85827d1981c2c565c627dd506afc0bdd11046368e8b71270e956563e34092336"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:40:18.759403Z","signature_b64":"Xu8XRt5ZwrBacqKpXMq9+tXXHHAYhVLQuSbP8CQoW4m15lKmnMNqCVN4F3soj4EwKL2kkbwxRRHhBSK2rHNnBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"afcb40fd1ab5d819f07e58d7a0704eb2b265ba7720a1d513630625bf7e1a1457","last_reissued_at":"2026-07-05T10:40:18.758900Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:40:18.758900Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SlowFast-LLaVA-1.5: A Family of Token-Efficient Video Large Language Models for Long-Form Video Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Afshin Dehghan, Jiasen Lu, Kai Kang, Meng Cao, Mingfei Gao, Mingze Xu, Shiyu Li, Yinfei Yang, Zhe Gan, Zhengfeng Lai","submitted_at":"2025-03-24T17:59:07Z","abstract_excerpt":"We introduce SlowFast-LLaVA-1.5 (abbreviated as SF-LLaVA-1.5), a family of video large language models (LLMs) offering a token-efficient solution for long-form video understanding. We incorporate the two-stream SlowFast mechanism into a streamlined training pipeline, and perform joint video-image training on a carefully curated data mixture of only publicly available datasets. Our primary focus is on highly efficient model scales (1B and 3B), demonstrating that even relatively small Video LLMs can achieve state-of-the-art performance on video understanding, meeting the demand for mobile-friend"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.18943","kind":"arxiv","version":2},"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/2503.18943/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":"2503.18943","created_at":"2026-07-05T10:40:18.758960+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.18943v2","created_at":"2026-07-05T10:40:18.758960+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.18943","created_at":"2026-07-05T10:40:18.758960+00:00"},{"alias_kind":"pith_short_12","alias_value":"V7FUB7I2WXMB","created_at":"2026-07-05T10:40:18.758960+00:00"},{"alias_kind":"pith_short_16","alias_value":"V7FUB7I2WXMBT4D6","created_at":"2026-07-05T10:40:18.758960+00:00"},{"alias_kind":"pith_short_8","alias_value":"V7FUB7I2","created_at":"2026-07-05T10:40:18.758960+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26014","citing_title":"STORM: Internalized Modeling for Spatial-Temporal Reasoning in Video-Language Models","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27074","citing_title":"IPIBench: Evaluating Interactive Proactive Intelligence of MLLMs under Continuous Streams","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28229","citing_title":"VidPrism: Heterogeneous Mixture of Experts for Image-to-Video Transfer","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22678","citing_title":"Swift Sampling: Selecting Temporal Surprises via Taylor Series","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06809","citing_title":"LookWhen? Fast Video Recognition by Learning When, Where, and What to Compute","ref_index":54,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK","json":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK.json","graph_json":"https://pith.science/api/pith-number/V7FUB7I2WXMBT4D6LDL2A4COWK/graph.json","events_json":"https://pith.science/api/pith-number/V7FUB7I2WXMBT4D6LDL2A4COWK/events.json","paper":"https://pith.science/paper/V7FUB7I2"},"agent_actions":{"view_html":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK","download_json":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK.json","view_paper":"https://pith.science/paper/V7FUB7I2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.18943&json=true","fetch_graph":"https://pith.science/api/pith-number/V7FUB7I2WXMBT4D6LDL2A4COWK/graph.json","fetch_events":"https://pith.science/api/pith-number/V7FUB7I2WXMBT4D6LDL2A4COWK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK/action/storage_attestation","attest_author":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK/action/author_attestation","sign_citation":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK/action/citation_signature","submit_replication":"https://pith.science/pith/V7FUB7I2WXMBT4D6LDL2A4COWK/action/replication_record"}},"created_at":"2026-07-05T10:40:18.758960+00:00","updated_at":"2026-07-05T10:40:18.758960+00:00"}