{"paper":{"title":"Bridging the Pose-Semantic Gap: A Cascade Framework for Text-Based Person Anomaly Search","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A two-stage cascade filters candidates by skeletal pose then verifies semantics with a multi-agent squad to bridge the gap where different actions share similar structures.","cross_cats":["cs.MM"],"primary_cat":"cs.CV","authors_text":"Chuxin Wang, Guijin Luo, Sihang Cai, Tao Jin, Yixuan Tang, Zequn Xie, Zhou Zhao","submitted_at":"2026-04-25T12:53:15Z","abstract_excerpt":"Text-based person anomaly search retrieves specific behavioral events from surveillance archives using natural-language queries. Although recent pose-aware methods align geometric structures well, they face a fundamental Pose-Semantic Gap: semantically different actions can share similar skeletal geometries. While Multimodal Large Language Models (MLLMs) can reduce this ambiguity, using them for large-scale retrieval is computationally prohibitive. We propose the Structure-Semantic Decoupled Cascade (SSDC) framework, which decouples retrieval into two stages: (1) Structure-Aware Coarse Retriev"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose the Structure-Semantic Decoupled Cascade (SSDC) framework, which decouples retrieval into two stages: (1) Structure-Aware Coarse Retrieval... and (2) Detective Squad Interaction... Experiments on the PAB benchmark show that SSDC achieves state-of-the-art performance by balancing efficiency and semantic reasoning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That skeletal geometry provides a sufficiently reliable coarse filter for semantically distinct actions and that the multi-agent LLM verification stage can resolve remaining ambiguities accurately without introducing new errors or prohibitive latency.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SSDC is a two-stage cascade that first filters by skeletal similarity then applies a multi-agent LLM squad for semantic verification, achieving SOTA on the PAB benchmark for text-based person anomaly search.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A two-stage cascade filters candidates by skeletal pose then verifies semantics with a multi-agent squad to bridge the gap where different actions share similar structures.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1499d35b1977e6d0d6e1d63ef7e5217b7be59dc48420a8d8fde366dbd62da02d"},"source":{"id":"2604.23282","kind":"arxiv","version":2},"verdict":{"id":"7a6fe0ff-1d84-4ad9-a15b-ec4b2e4e5ba1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T08:24:13.323832Z","strongest_claim":"We propose the Structure-Semantic Decoupled Cascade (SSDC) framework, which decouples retrieval into two stages: (1) Structure-Aware Coarse Retrieval... and (2) Detective Squad Interaction... Experiments on the PAB benchmark show that SSDC achieves state-of-the-art performance by balancing efficiency and semantic reasoning.","one_line_summary":"SSDC is a two-stage cascade that first filters by skeletal similarity then applies a multi-agent LLM squad for semantic verification, achieving SOTA on the PAB benchmark for text-based person anomaly search.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That skeletal geometry provides a sufficiently reliable coarse filter for semantically distinct actions and that the multi-agent LLM verification stage can resolve remaining ambiguities accurately without introducing new errors or prohibitive latency.","pith_extraction_headline":"A two-stage cascade filters candidates by skeletal pose then verifies semantics with a multi-agent squad to bridge the gap where different actions share similar structures."},"integrity":{"clean":false,"summary":{"advisory":1,"critical":0,"by_detector":{"doi_compliance":{"total":1,"advisory":1,"critical":0,"informational":0}},"informational":0},"endpoint":"/pith/2604.23282/integrity.json","findings":[{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.23919/cje.2025.00.215Towards) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":9,"audited_at":"2026-05-19T23:16:39.952796Z","detected_doi":"10.23919/cje.2025.00.215Towards","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T09:36:08.064601Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T23:16:39.952796Z","status":"completed","version":"1.0.0","findings_count":1}],"snapshot_sha256":"ce7d2ea1dc2a6dc90b2212a582a296c570bab740de27697c4d494b03bbcf1ed7"},"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"}