{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:4KXQSEKCACNIJPCAS2CCXQUMFW","short_pith_number":"pith:4KXQSEKC","schema_version":"1.0","canonical_sha256":"e2af091142009a84bc4096842bc28c2db62a356946ac6d3cd57d974549e11f2f","source":{"kind":"arxiv","id":"2605.13667","version":1},"attestation_state":"computed","paper":{"title":"SceneGraphVLM: Dynamic Scene Graph Generation from Video with Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SceneGraphVLM generates complete scene graphs from videos in about one second using compact vision-language models and targeted reinforcement learning.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dmitry Yudin, Mark Gizetdinov, Vladislav Makarov","submitted_at":"2026-05-13T15:27:41Z","abstract_excerpt":"Scene graph generation provides a compact structured representation for visual perception, but accurate and fast graph prediction from images and videos remains challenging. Recent VLM-based methods can generate scene graphs end-to-end as structured text, yet often produce long outputs with irrelevant objects and relations. We present SceneGraphVLM, a compact method for image and video scene graph generation with small visual language models. SceneGraphVLM serializes graphs in a token-efficient TOON format and trains the model in two stages: supervised fine-tuning followed by reinforcement lea"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.13667","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T15:27:41Z","cross_cats_sorted":[],"title_canon_sha256":"f04ec2be006a9833106289e9936921d6238aa3883a63199fe5bc8268919a8873","abstract_canon_sha256":"d07ffdb47a3c956d7a68e5aaa76b132c5c7815a7172b31fcc35853cc1d19a574"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:17.232626Z","signature_b64":"Mwt/Iottin8e3OyE3+V6hqeQjTr8OsOAV5RDpBZFi7/uT3pDp78L/3wD9Cfs4CgVbuAGR74eO+rJyr9u/uj1Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e2af091142009a84bc4096842bc28c2db62a356946ac6d3cd57d974549e11f2f","last_reissued_at":"2026-05-18T02:44:17.232036Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:17.232036Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SceneGraphVLM: Dynamic Scene Graph Generation from Video with Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SceneGraphVLM generates complete scene graphs from videos in about one second using compact vision-language models and targeted reinforcement learning.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dmitry Yudin, Mark Gizetdinov, Vladislav Makarov","submitted_at":"2026-05-13T15:27:41Z","abstract_excerpt":"Scene graph generation provides a compact structured representation for visual perception, but accurate and fast graph prediction from images and videos remains challenging. Recent VLM-based methods can generate scene graphs end-to-end as structured text, yet often produce long outputs with irrelevant objects and relations. We present SceneGraphVLM, a compact method for image and video scene graph generation with small visual language models. SceneGraphVLM serializes graphs in a token-efficient TOON format and trains the model in two stages: supervised fine-tuning followed by reinforcement lea"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"With compact VLMs and vLLM-accelerated decoding, SceneGraphVLM achieves a strong quality-speed trade-off, improves precision-oriented SGG metrics while preserving reasonable recall, and generates complete scene graphs with approximately one-second latency.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the hallucination-aware RL rewards successfully balance coverage and precision on the target benchmarks without introducing new failure modes or requiring dataset-specific tuning that does not generalize.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SceneGraphVLM generates dynamic scene graphs from video using compact VLMs, TOON serialization, and hallucination-aware RL to improve precision and achieve one-second latency.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SceneGraphVLM generates complete scene graphs from videos in about one second using compact vision-language models and targeted reinforcement learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"cdc73857edbf736eec9311abebb33e553a65960994222df0d88be1ee398d2d86"},"source":{"id":"2605.13667","kind":"arxiv","version":1},"verdict":{"id":"2ca0be25-326a-43c7-8530-f666df097943","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T20:36:53.643217Z","strongest_claim":"With compact VLMs and vLLM-accelerated decoding, SceneGraphVLM achieves a strong quality-speed trade-off, improves precision-oriented SGG metrics while preserving reasonable recall, and generates complete scene graphs with approximately one-second latency.","one_line_summary":"SceneGraphVLM generates dynamic scene graphs from video using compact VLMs, TOON serialization, and hallucination-aware RL to improve precision and achieve one-second latency.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the hallucination-aware RL rewards successfully balance coverage and precision on the target benchmarks without introducing new failure modes or requiring dataset-specific tuning that does not generalize.","pith_extraction_headline":"SceneGraphVLM generates complete scene graphs from videos in about one second using compact vision-language models and targeted reinforcement learning."},"references":{"count":48,"sample":[{"doi":"","year":2025,"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","ref_index":1,"cited_arxiv_id":"2511.21631","is_internal_anchor":true},{"doi":"","year":null,"title":"Qwen2.5-vl technical report","work_id":"8cce8101-1a0a-4da5-a8d4-8b72b9d0c60f","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","ref_index":3,"cited_arxiv_id":"2502.13923","is_internal_anchor":true},{"doi":"","year":2024,"title":"Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scaling","work_id":"ee70bdc8-4656-4849-ada7-ce42a2278d70","ref_index":4,"cited_arxiv_id":"2412.05271","is_internal_anchor":true},{"doi":"","year":2025,"title":"Compile scene graphs with reinforcement learning","work_id":"bb1a8a98-c413-4df6-8e94-70284c64162f","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":48,"snapshot_sha256":"97a2867af96301a2557f855dc219193bd504a966b68631f2e0dc63e3ebe7d690","internal_anchors":11},"formal_canon":{"evidence_count":2,"snapshot_sha256":"6a5a90d487e236227f8f97570d3224441a144ba0a757d583bc544af44218eb26"},"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":"2605.13667","created_at":"2026-05-18T02:44:17.232130+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13667v1","created_at":"2026-05-18T02:44:17.232130+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13667","created_at":"2026-05-18T02:44:17.232130+00:00"},{"alias_kind":"pith_short_12","alias_value":"4KXQSEKCACNI","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"4KXQSEKCACNIJPCA","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"4KXQSEKC","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW","json":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW.json","graph_json":"https://pith.science/api/pith-number/4KXQSEKCACNIJPCAS2CCXQUMFW/graph.json","events_json":"https://pith.science/api/pith-number/4KXQSEKCACNIJPCAS2CCXQUMFW/events.json","paper":"https://pith.science/paper/4KXQSEKC"},"agent_actions":{"view_html":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW","download_json":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW.json","view_paper":"https://pith.science/paper/4KXQSEKC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13667&json=true","fetch_graph":"https://pith.science/api/pith-number/4KXQSEKCACNIJPCAS2CCXQUMFW/graph.json","fetch_events":"https://pith.science/api/pith-number/4KXQSEKCACNIJPCAS2CCXQUMFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW/action/storage_attestation","attest_author":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW/action/author_attestation","sign_citation":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW/action/citation_signature","submit_replication":"https://pith.science/pith/4KXQSEKCACNIJPCAS2CCXQUMFW/action/replication_record"}},"created_at":"2026-05-18T02:44:17.232130+00:00","updated_at":"2026-05-18T02:44:17.232130+00:00"}