{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RQSG5Y2B2BFPEOI2WVKSOXBBAT","short_pith_number":"pith:RQSG5Y2B","schema_version":"1.0","canonical_sha256":"8c246ee341d04af2391ab555275c2104d00a3361cc446a42c898241701b23b4d","source":{"kind":"arxiv","id":"2605.15320","version":1},"attestation_state":"computed","paper":{"title":"FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A feed-forward model reconstructs animatable 3D Gaussian head avatars from few unposed photos in seconds without per-subject optimization.","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.GR","authors_text":"Gordon Guocheng Qian, Hao Li, Jiahao Luo, Jian Wang, Thuan Hoang Nguyen, Yinyu Nie","submitted_at":"2026-05-14T18:33:49Z","abstract_excerpt":"Avatar reconstruction has traditionally relied on per-subject optimization that requires hours of computation or on expensive preprocessing that limits scalability. We introduce FFAvatar, a generalizable feed-forward framework that reconstructs high-quality, animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds. FFAvatar fuses information from multiple source images into a unified canonical Gaussian representation through Multi-View Query-Former, which is animated via FLAME parameters predicted end-to-end directly from pixels, eliminating the overhead of offline "},"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":"2605.15320","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.GR","submitted_at":"2026-05-14T18:33:49Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"82bf35c1b3d769c5d7f62f5f4aad2e8ac5fbdf1da4a22dffce554e20ca733a23","abstract_canon_sha256":"ec498ea43cd1d203c2d4451c9c56041b3c2d507434d90a9db151e2a063ea244b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:52.480880Z","signature_b64":"MEMm84Tcb7GNVwa1EoC/SV5ZLYL2KSlhNLdkSKBTaSbKuzUnPTARdNPtyFZE+R8qrXkh8a+UKIuYZVi9We44DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c246ee341d04af2391ab555275c2104d00a3361cc446a42c898241701b23b4d","last_reissued_at":"2026-05-20T00:00:52.480050Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:52.480050Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FFAvatar: Few-Shot, Feed-Forward, and Generalizable Avatar Reconstruction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A feed-forward model reconstructs animatable 3D Gaussian head avatars from few unposed photos in seconds without per-subject optimization.","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.GR","authors_text":"Gordon Guocheng Qian, Hao Li, Jiahao Luo, Jian Wang, Thuan Hoang Nguyen, Yinyu Nie","submitted_at":"2026-05-14T18:33:49Z","abstract_excerpt":"Avatar reconstruction has traditionally relied on per-subject optimization that requires hours of computation or on expensive preprocessing that limits scalability. We introduce FFAvatar, a generalizable feed-forward framework that reconstructs high-quality, animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds. FFAvatar fuses information from multiple source images into a unified canonical Gaussian representation through Multi-View Query-Former, which is animated via FLAME parameters predicted end-to-end directly from pixels, eliminating the overhead of offline "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"FFAvatar reconstructs high-quality, animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds and sets a new standard for identity preservation, geometric consistency, and animation fidelity, outperforming the state-of-the-art LAM by a substantial 5.5 PSNR gain on the NeRSemble benchmark.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The three-stage training curriculum on monocular video data with over 1M identities followed by multi-view fine-tuning produces priors that generalize to arbitrary few-shot unposed inputs without requiring offline pose or FLAME extraction.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FFAvatar is a generalizable feed-forward framework that reconstructs high-quality animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds via Multi-View Query-Former and end-to-end FLAME prediction.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A feed-forward model reconstructs animatable 3D Gaussian head avatars from few unposed photos in seconds without per-subject optimization.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"75e34c773e84d4dd26aa2d521303511d5c781e708d8716837205a425aa8f532a"},"source":{"id":"2605.15320","kind":"arxiv","version":1},"verdict":{"id":"e4e85eff-722b-483f-9963-b012e2fd7693","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T15:35:09.932931Z","strongest_claim":"FFAvatar reconstructs high-quality, animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds and sets a new standard for identity preservation, geometric consistency, and animation fidelity, outperforming the state-of-the-art LAM by a substantial 5.5 PSNR gain on the NeRSemble benchmark.","one_line_summary":"FFAvatar is a generalizable feed-forward framework that reconstructs high-quality animatable 3D Gaussian head avatars from few-shot unposed portrait images in seconds via Multi-View Query-Former and end-to-end FLAME prediction.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The three-stage training curriculum on monocular video data with over 1M identities followed by multi-view fine-tuning produces priors that generalize to arbitrary few-shot unposed inputs without requiring offline pose or FLAME extraction.","pith_extraction_headline":"A feed-forward model reconstructs animatable 3D Gaussian head avatars from few unposed photos in seconds without per-subject optimization."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15320/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:01:18.151828Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:41:05.371183Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.209702Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.769391Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"8b1737c078db627f9c72437b31acb22be6c03d33856edd8ad907455dc62826ec"},"references":{"count":33,"sample":[{"doi":"","year":2023,"title":"A morphable model for the synthesis of 3d faces.Seminal Graphics Papers: Pushing the Boundaries, Volume 2, pages 157–164, 2023","work_id":"4de83edc-46a9-477f-9a2a-8313493f5309","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Generalizable and animatable gaussian head avatar","work_id":"05900833-719b-4725-a2bf-ac1cbe713d72","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"GPA- vatar: Generalizable and precise head avatar from image(s)","work_id":"3222b022-86ea-4e84-9e60-b5593393f9e8","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Black, and Timo Bolkart","work_id":"9fa5c3de-2ba5-4248-842d-1da9f10d0f01","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"Arcface: Additive angular margin loss for deep face recognition","work_id":"466a3f28-f5e8-456f-9a6a-a6b0fdb10a43","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":33,"snapshot_sha256":"959e3f5bf114f0af6163f2b061a383e6b15eb1948554f6f878ea580eeeb843c7","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"4d649faea72ef4155c8dd7c35325a62f0e8e922fac135dcd385a1fd6fe99e05f"},"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":"2605.15320","created_at":"2026-05-20T00:00:52.480179+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15320v1","created_at":"2026-05-20T00:00:52.480179+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15320","created_at":"2026-05-20T00:00:52.480179+00:00"},{"alias_kind":"pith_short_12","alias_value":"RQSG5Y2B2BFP","created_at":"2026-05-20T00:00:52.480179+00:00"},{"alias_kind":"pith_short_16","alias_value":"RQSG5Y2B2BFPEOI2","created_at":"2026-05-20T00:00:52.480179+00:00"},{"alias_kind":"pith_short_8","alias_value":"RQSG5Y2B","created_at":"2026-05-20T00:00:52.480179+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":1,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT","json":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT.json","graph_json":"https://pith.science/api/pith-number/RQSG5Y2B2BFPEOI2WVKSOXBBAT/graph.json","events_json":"https://pith.science/api/pith-number/RQSG5Y2B2BFPEOI2WVKSOXBBAT/events.json","paper":"https://pith.science/paper/RQSG5Y2B"},"agent_actions":{"view_html":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT","download_json":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT.json","view_paper":"https://pith.science/paper/RQSG5Y2B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15320&json=true","fetch_graph":"https://pith.science/api/pith-number/RQSG5Y2B2BFPEOI2WVKSOXBBAT/graph.json","fetch_events":"https://pith.science/api/pith-number/RQSG5Y2B2BFPEOI2WVKSOXBBAT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT/action/storage_attestation","attest_author":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT/action/author_attestation","sign_citation":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT/action/citation_signature","submit_replication":"https://pith.science/pith/RQSG5Y2B2BFPEOI2WVKSOXBBAT/action/replication_record"}},"created_at":"2026-05-20T00:00:52.480179+00:00","updated_at":"2026-05-20T00:00:52.480179+00:00"}