{"paper":{"title":"ScaleMoGen: Autoregressive Next-Scale Prediction for Human Motion Generation","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"ScaleMoGen generates human motions by autoregressively predicting discrete tokens from coarse to fine skeletal-temporal scales.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bing Zhou, Chuan Guo, Hojun Jang, Inwoo Hwang, Jian Wang, Young Min Kim","submitted_at":"2026-05-12T07:58:58Z","abstract_excerpt":"We present ScaleMoGen, a scale-wise autoregressive framework for text-driven human motion generation. Unlike conventional autoregressive approaches that rely on standard next-token prediction, ScaleMoGen frames motion generation as a coarse-to-fine process. We quantize 3D motions into compositional discrete tokens across multiple skeletal-emporal scales of increasing granularity, learning to generate motion by autoregressively predicting next-scale token maps. To maintain structural integrity, our motion tokenizers and quantizers are explicitly designed so that discrete tokens at every scale s"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ScaleMoGen achieves state-of-the-art performance, establishing an FID of 0.030 (vs. 0.045 for MoMask) on HumanML3D and a CLIP Score of 0.693 (vs. 0.685 for MoMask++) on the SnapMoGen dataset, while enabling training-free text-guided motion editing.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That quantizing 3D motions into compositional discrete tokens across multiple skeletal-temporal scales strictly preserves the skeletal hierarchy and that bitwise quantization stabilizes optimization for high-detail motions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ScaleMoGen introduces a scale-wise autoregressive framework that quantizes motions into hierarchical discrete tokens and predicts next-scale maps to achieve SOTA FID 0.030 on HumanML3D and text-guided editing.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ScaleMoGen generates human motions by autoregressively predicting discrete tokens from coarse to fine skeletal-temporal scales.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"86cfba83bde1190a64f5e13bb3c6c4e35c18503b54c4952deaed972502ca7d50"},"source":{"id":"2605.11704","kind":"arxiv","version":2},"verdict":{"id":"0f5be8e1-0bb7-4cd7-8d8b-b28ce2498e6f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T06:20:11.668434Z","strongest_claim":"ScaleMoGen achieves state-of-the-art performance, establishing an FID of 0.030 (vs. 0.045 for MoMask) on HumanML3D and a CLIP Score of 0.693 (vs. 0.685 for MoMask++) on the SnapMoGen dataset, while enabling training-free text-guided motion editing.","one_line_summary":"ScaleMoGen introduces a scale-wise autoregressive framework that quantizes motions into hierarchical discrete tokens and predicts next-scale maps to achieve SOTA FID 0.030 on HumanML3D and text-guided editing.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That quantizing 3D motions into compositional discrete tokens across multiple skeletal-temporal scales strictly preserves the skeletal hierarchy and that bitwise quantization stabilizes optimization for high-detail motions.","pith_extraction_headline":"ScaleMoGen generates human motions by autoregressively predicting discrete tokens from coarse to fine skeletal-temporal scales."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.11704/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-21T00:01:32.042021Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T13:55:12.273920Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-20T03:42:00.474062Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T11:39:30.464506Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"d46587df6bbbe9ce7f3ebc65ec28a458c45c6a38c1d9e831faa7a5e1036bfc0e"},"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"}