{"paper":{"title":"Frequency-Enhanced Diffusion Models: Curriculum-Guided Semantic Alignment for Zero-Shot Skeleton Action Recognition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Frequency-aware diffusion models recover fine-grained motion details for zero-shot skeleton action recognition.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Jingyu Pan, Yuxi Zhou, Zhengbo Zhang, Zhigang Tu, Zhiyu Lin","submitted_at":"2026-04-10T07:42:47Z","abstract_excerpt":"Human action recognition is pivotal in computer vision, with applications ranging from surveillance to human-robot interaction. Despite the effectiveness of supervised skeleton-based methods, their reliance on exhaustive annotation limits generalization to novel actions. Zero-Shot Skeleton Action Recognition (ZSAR) emerges as a promising paradigm, yet it faces challenges due to the spectral bias of diffusion models, which oversmooth high-frequency dynamics. Here, we propose Frequency-Aware Diffusion for Skeleton-Text Matching (FDSM), integrating a Semantic-Guided Spectral Residual Module, a Ti"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our approach effectively recovers fine-grained motion details, achieving state-of-the-art performance on NTU RGB+D, PKU-MMD, and Kinetics-skeleton datasets.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the spectral bias of diffusion models is the primary bottleneck in zero-shot skeleton action recognition and that the three proposed modules (Semantic-Guided Spectral Residual Module, Timestep-Adaptive Spectral Loss, Curriculum-based Semantic Abstraction) directly correct it without introducing compensating errors or requiring dataset-specific tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FDSM recovers fine-grained motion details in zero-shot skeleton action recognition by integrating semantic-guided spectral residual, timestep-adaptive spectral loss, and curriculum-based semantic abstraction, reaching state-of-the-art on NTU RGB+D, PKU-MMD, and Kinetics-skeleton.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Frequency-aware diffusion models recover fine-grained motion details for zero-shot skeleton action recognition.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"fe89374dda37bff27fcd35277880ab25902691fdde3307236e58bbf2b68e06e4"},"source":{"id":"2604.09063","kind":"arxiv","version":3},"verdict":{"id":"44936811-1bfa-4b6f-9e5b-83b3fe7b60d1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T16:51:09.040451Z","strongest_claim":"Our approach effectively recovers fine-grained motion details, achieving state-of-the-art performance on NTU RGB+D, PKU-MMD, and Kinetics-skeleton datasets.","one_line_summary":"FDSM recovers fine-grained motion details in zero-shot skeleton action recognition by integrating semantic-guided spectral residual, timestep-adaptive spectral loss, and curriculum-based semantic abstraction, reaching state-of-the-art on NTU RGB+D, PKU-MMD, and Kinetics-skeleton.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the spectral bias of diffusion models is the primary bottleneck in zero-shot skeleton action recognition and that the three proposed modules (Semantic-Guided Spectral Residual Module, Timestep-Adaptive Spectral Loss, Curriculum-based Semantic Abstraction) directly correct it without introducing compensating errors or requiring dataset-specific tuning.","pith_extraction_headline":"Frequency-aware diffusion models recover fine-grained motion details for zero-shot skeleton action recognition."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.09063/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"}