{"paper":{"title":"FCBV-Net: Category-Level Robotic Garment Smoothing via Feature-Conditioned Bimanual Value Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Conditioning bimanual action values on frozen 3D geometric features lets robots smooth unseen garments with far smaller performance loss than baselines.","cross_cats":["cs.AI"],"primary_cat":"cs.RO","authors_text":"Jing Qiu, Mohammed Daba","submitted_at":"2025-08-07T08:37:45Z","abstract_excerpt":"Category-level generalization for robotic garment manipulation, such as bimanual smoothing, remains a significant hurdle due to high dimensionality, complex dynamics, and intra-category variations. Current approaches often struggle, either overfitting with concurrently learned visual features for a specific instance or, despite Category-level perceptual generalization, failing to predict the value of synergistic bimanual actions. We propose the Feature-Conditioned bimanual Value Network (FCBV-Net), operating on 3D point clouds to specifically enhance category-level policy generalization for ga"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"the decoupling of geometric understanding from bimanual action value learning enables better category-level generalization","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That pre-trained and frozen dense geometric features extracted from 3D point clouds remain sufficiently informative and robust across intra-category variations in the CLOTH3D dataset without any task-specific fine-tuning or adaptation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FCBV-Net achieves superior category-level generalization for bimanual garment smoothing by conditioning value prediction on static pre-trained dense geometric features from point clouds, showing only 11.5% efficiency drop on unseen garments versus much larger drops in baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Conditioning bimanual action values on frozen 3D geometric features lets robots smooth unseen garments with far smaller performance loss than baselines.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"580b48da755ffd9868b295002f46a5b1438f43f1cb9ada26b62770067be1470a"},"source":{"id":"2508.05153","kind":"arxiv","version":2},"verdict":{"id":"a401bd7d-83ce-4f72-955d-25dc0f4bd198","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T01:12:13.308879Z","strongest_claim":"the decoupling of geometric understanding from bimanual action value learning enables better category-level generalization","one_line_summary":"FCBV-Net achieves superior category-level generalization for bimanual garment smoothing by conditioning value prediction on static pre-trained dense geometric features from point clouds, showing only 11.5% efficiency drop on unseen garments versus much larger drops in baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That pre-trained and frozen dense geometric features extracted from 3D point clouds remain sufficiently informative and robust across intra-category variations in the CLOTH3D dataset without any task-specific fine-tuning or adaptation.","pith_extraction_headline":"Conditioning bimanual action values on frozen 3D geometric features lets robots smooth unseen garments with far smaller performance loss than baselines."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.05153/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":27,"sample":[{"doi":"","year":2014,"title":"Ide ntifying the potential for robotics to assist older adults in differe nt living environments","work_id":"db79ed13-822e-4656-a2d3-7a2283fbbdb3","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2008,"title":"Managing activity difficulties at ho me: A survey of medicare beneficiaries","work_id":"4d3f5fc8-061c-470c-8165-ddc3977443fe","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Challenges and outlook in robotic manipula tion of deformable objects","work_id":"9fb447fa-4a8f-47e8-ab39-80188cc8d38c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Unigarmentman ip: A unified framework for category-level garment manipulatio n via dense visual correspondence","work_id":"ce43107d-4c31-4e91-b754-667d13f1a000","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Speedfolding: Learning efficient bimanual folding of garm ents","work_id":"9b226935-fd30-4c0b-84a9-4992c2a7ae68","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":27,"snapshot_sha256":"9c774b658d248ce63a1a63b35474ecdc9472c8d160a37c8013cc56dc68616dee","internal_anchors":1},"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"}