{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RLKEZFWWC4ROM2B34EC2AKKJ7W","short_pith_number":"pith:RLKEZFWW","schema_version":"1.0","canonical_sha256":"8ad44c96d61722e6683be105a02949fdabef950933a3868cc1fad99bb98e69e9","source":{"kind":"arxiv","id":"2412.00161","version":2},"attestation_state":"computed","paper":{"title":"STEP: Enhancing Video-LLMs' Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Haiyi Qiu, Juncheng Li, Kaihang Pan, Long Qian, Minghe Gao, Qifan Yu, Siliang Tang, Tat-Seng Chua, Wenjie Wang, Yueting Zhuang","submitted_at":"2024-11-29T11:54:55Z","abstract_excerpt":"Video Large Language Models (Video-LLMs) have recently shown strong performance in basic video understanding tasks, such as captioning and coarse-grained question answering, but struggle with compositional reasoning that requires multi-step spatio-temporal inference across object relations, interactions, and events. The hurdles to enhancing this capability include extensive manual labor, the lack of spatio-temporal compositionality in existing data and the absence of explicit reasoning supervision. In this paper, we propose STEP, a novel graph-guided self-training method that enables Video-LLM"},"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":"2412.00161","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-11-29T11:54:55Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a06473a087928fb0f3e23f654dd7def3943ee51260a139d5e5c8acf6a6fceeee","abstract_canon_sha256":"8d467d18113208aaa29a6d0bcf148447bde3a40095f21d8c383eec3dd6807437"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:18.039758Z","signature_b64":"EiMFRmhSR37JE6pmZam3FXOe6E5lX/tr/zuRbv1zD/C8jMoNCtZ6aF3Lcw4HwU9g9It1SYSsOejksrEaXsGGCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8ad44c96d61722e6683be105a02949fdabef950933a3868cc1fad99bb98e69e9","last_reissued_at":"2026-07-05T10:41:18.039226Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:18.039226Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"STEP: Enhancing Video-LLMs' Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Haiyi Qiu, Juncheng Li, Kaihang Pan, Long Qian, Minghe Gao, Qifan Yu, Siliang Tang, Tat-Seng Chua, Wenjie Wang, Yueting Zhuang","submitted_at":"2024-11-29T11:54:55Z","abstract_excerpt":"Video Large Language Models (Video-LLMs) have recently shown strong performance in basic video understanding tasks, such as captioning and coarse-grained question answering, but struggle with compositional reasoning that requires multi-step spatio-temporal inference across object relations, interactions, and events. The hurdles to enhancing this capability include extensive manual labor, the lack of spatio-temporal compositionality in existing data and the absence of explicit reasoning supervision. In this paper, we propose STEP, a novel graph-guided self-training method that enables Video-LLM"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00161","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/2412.00161/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":"2412.00161","created_at":"2026-07-05T10:41:18.039287+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.00161v2","created_at":"2026-07-05T10:41:18.039287+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00161","created_at":"2026-07-05T10:41:18.039287+00:00"},{"alias_kind":"pith_short_12","alias_value":"RLKEZFWWC4RO","created_at":"2026-07-05T10:41:18.039287+00:00"},{"alias_kind":"pith_short_16","alias_value":"RLKEZFWWC4ROM2B3","created_at":"2026-07-05T10:41:18.039287+00:00"},{"alias_kind":"pith_short_8","alias_value":"RLKEZFWW","created_at":"2026-07-05T10:41:18.039287+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05736","citing_title":"VTI-CoT: Visual-Textual Interleaved Chain of Thought for Video Reasoning","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W","json":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W.json","graph_json":"https://pith.science/api/pith-number/RLKEZFWWC4ROM2B34EC2AKKJ7W/graph.json","events_json":"https://pith.science/api/pith-number/RLKEZFWWC4ROM2B34EC2AKKJ7W/events.json","paper":"https://pith.science/paper/RLKEZFWW"},"agent_actions":{"view_html":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W","download_json":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W.json","view_paper":"https://pith.science/paper/RLKEZFWW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.00161&json=true","fetch_graph":"https://pith.science/api/pith-number/RLKEZFWWC4ROM2B34EC2AKKJ7W/graph.json","fetch_events":"https://pith.science/api/pith-number/RLKEZFWWC4ROM2B34EC2AKKJ7W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W/action/storage_attestation","attest_author":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W/action/author_attestation","sign_citation":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W/action/citation_signature","submit_replication":"https://pith.science/pith/RLKEZFWWC4ROM2B34EC2AKKJ7W/action/replication_record"}},"created_at":"2026-07-05T10:41:18.039287+00:00","updated_at":"2026-07-05T10:41:18.039287+00:00"}