{"id":"ee93be27-4509-4faa-93fa-f73db98a571a","arxiv_id":"2607.05938","paper_version":1,"verdict":"CONDITIONAL","confidence":"LOW","novelty_score":5.5,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"Prior-first body-hand kinematic model with layered adapters for real-time, low-supervision hand motion completion conditioned on body and semantics.","lead":"A prior-first framework learns body-hand kinematics from unlabeled motion, then adds light adapters for semantic hand control. It aims at real-time, controllable hand completion for animation with little labeled data.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.5","headline":"Frozen prior + light adapters may not preserve body-hand coupling under semantic control","rationale":"The paper is a clear systems contribution: learn kinematics first from unlabeled data, then add cheap semantic adapters. That design is plausible and the strongest claim is empirical (better plausibility/robustness/controllability vs end-to-end, especially low-resource and cross-dataset, plus real-time). The single load-bearing soft spot is exactly the reader's weakest_assumption: that freezing the prior and controlling only through light, layered adapters is sufficient for semantic control without relearning or breaking body-hand coupling. I do not see a stronger internal inconsistency or a more central missing condition from the available argument. Circularity is low for an empirical comparison paper; novelty is moderate (prior/adapter pattern applied carefully). No machine-checked proofs or parameter-free claims to reweight. Verdict stays CONDITIONAL: accept-shaped if the experiments hold and the coupling is preserved under control; code/data and a full empirical audit remain open. Confidence remains appropriately low without independent re-run. Agreement with the reader is full on the load-bearing concern; no verdict shift is warranted.","tokens_in":2128,"tokens_out":611,"duration_ms":8024,"concrete_test":"Re-run the main kinematic-plausibility metrics (e.g., hand-body coordination / penetration / joint-limit violations from the evaluation tables) on the same held-out sequences under (i) frozen prior alone, (ii) prior + self-supervised attribute adapters, and (iii) prior + text adapters, with body motion held fixed. If (ii) or (iii) degrade coordination metrics by more than ~10% relative to (i) while improving semantic metrics, the claim that adapters leave the prior's coupling intact fails and the low-resource advantage needs re-scoping.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim rests on freezing a streaming autoregressive body-hand kinematic prior (learned from unlabeled motion) and injecting control only via lightweight, semantically-layered adapters at chosen kinematic levels. For this to deliver both kinematic plausibility and semantic controllability without end-to-end joint training, the adapters must not break the prior's learned mechanical body-hand coupling. The paper's methods and low-supervision claims (abstract; adapter sections) treat this as given: limited labeled data and self-supervised attributes can drive controllable hands while the frozen prior continues to enforce coordination. That is the least secure condition. If adapters at higher semantic layers effectively override or dilute the prior's coupling (especially under text control or cross-dataset body dynamics), the reported gains in plausibility and robustness vs end-to-end baselines would be overstated or regime-specific. The reader's weakest_assumption correctly flags this; the load-bearing risk is whether the frozen-prior + adapter design actually keeps kinematics intact when control is applied, not merely whether adapters can be trained.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.5","summary":"The manuscript proposes a prior-first, condition-second framework for body-conditioned hand motion completion. A streaming autoregressive body-hand kinematic prior is first trained on large-scale unlabeled motion to capture intrinsic body-hand coordination and mechanical coupling. Semantic control is then added via lightweight, semantically-layered adapters on the frozen prior, supporting self-supervised attribute control and weakly supervised text-driven control with only a few hours of labeled data. The authors claim improved kinematic plausibility, robustness, and controllability relative to end-to-end conditioned baselines, especially in low-resource and cross-dataset regimes, together with real-time inference and an interactive authoring workflow for production animation.","tokens_in":2304,"tokens_out":1069,"duration_ms":41657,"significance":"If the empirical claims hold, the work offers a modular and practical alternative to end-to-end conditioned hand motion models. Decoupling a reusable kinematic prior from task-specific control interfaces is valuable for animation pipelines where labeled semantic data are scarce and streaming real-time generation is required. Credit is due for the streaming autoregressive prior design, the explicit focus on low-supervision controllability, the cross-dataset robustness emphasis, and the interactive authoring demonstration. These elements, if validated, constitute a useful design pattern for controllable body-hand synthesis under limited labels.","major_comments":[{"comment":"The central claim that freezing the body-hand kinematic prior and injecting control only via lightweight semantically-layered adapters preserves mechanical body-hand coupling is load-bearing for the reported gains in kinematic plausibility and robustness versus end-to-end baselines. The manuscript should provide direct evidence—e.g., body-hand joint correlation or relative-pose consistency metrics with vs. without adapters, under attribute and text control, and in cross-dataset settings—that higher-layer adapters do not dilute or override the prior’s learned coordination. Without such measurements, the advantage of the frozen-prior design remains incompletely substantiated relative to the weakest design assumption.","section":"Methods (prior + semantically-layered adapters); abstract claims on coupling"},{"comment":"The low-resource and “few hours of labeled data” claims are central to the condition-second contribution. Please report exact labeled data volumes (hours/frames) for each adapter setting, the corresponding training budgets and data scales for end-to-end baselines, and a controlled ablation that varies labeled-data scale. Without matched budgets and a scale curve, the claimed efficiency and robustness advantages over end-to-end conditioning cannot be fairly assessed.","section":"Experiments / low-resource and weakly supervised text control"},{"comment":"Kinematic plausibility is multi-faceted. Clarify which metrics (hand-body relative pose error, joint-limit violations, interpenetration, temporal jerk/acceleration consistency, or analogous coordination scores) drive the “improved kinematic plausibility” claim, and whether they are evaluated under free semantic control as well as pure body-conditioned completion. If primary tables emphasize reconstruction-style errors under body conditioning alone, they do not fully address coupling preservation when adapters inject strong semantic signals.","section":"Evaluation metrics and main result tables"}],"minor_comments":[{"comment":"Define adapter placement and “kinematic levels” more precisely (which layers/joints receive which semantic signals) and state adapter rank/capacity relative to the frozen prior so that “lightweight” is reproducible.","section":"Adapter architecture subsection"},{"comment":"Related work should more explicitly position the design against frozen-backbone + adapter literature in motion and vision/language, and against prior body-hand or hand-only completion models, to clarify what is new beyond the prior-first framing.","section":"Related Work"},{"comment":"For the interactive authoring and real-time claims, report latency, streaming context length, and hardware so that production applicability can be judged independently of qualitative demos.","section":"Real-time inference / authoring workflow"},{"comment":"Ensure all baselines are described with architecture, training data, and conditioning interface parity; any mismatch should be stated explicitly in the experimental protocol.","section":"Baselines and experimental setup"},{"comment":"Minor presentation: expand acronyms on first use, check figure axis labels and caption self-containment, and verify that the homepage link and any supplementary videos are stable for review.","section":"Throughout / figures"}],"recommendation":"major_revision","confidential_remarks":"The load-bearing risk is the frozen-prior + adapter coupling assumption flagged in the stress-test note; I treat it as a request for evidence rather than a demonstrated failure. Full experimental tables and ablations were not independently re-run here. The manuscript fits cs.GR / character animation scope. Novelty relative to the broader adapter literature should be watched in revision so the contribution is not overstated as purely architectural."},"author_rebuttal":{"model":"grok-4.5","summary":"We thank the referee for a careful and constructive review. The three major comments correctly identify load-bearing claims of the prior-first, condition-second design: (i) that freezing the body–hand kinematic prior and injecting control only via lightweight semantically-layered adapters preserves mechanical coupling; (ii) that the low-resource / few-hours labeled-data claims require matched budgets and a scale curve; and (iii) that “improved kinematic plausibility” must be tied to explicit coordination metrics under free semantic control, not only reconstruction under body conditioning. We address each point below with concrete revisions: new coupling-preservation metrics (with vs. without adapters, under attribute and text control, and cross-dataset), exact labeled volumes and matched end-to-end budgets plus a labeled-data scale ablation, and a clarified metric suite with primary-table updates that evaluate coordination under free semantic control. We believe these changes fully substantiate the design claims and make the empirical case fair and transparent.","responses":[{"response":"We agree this is load-bearing and that the original manuscript under-specified direct coupling evidence. We will add a dedicated “Coupling Preservation” analysis. Concretely: (1) body–hand joint correlation matrices and Pearson/Spearman scores between body root/upper-limb velocities and hand joint angles, comparing frozen prior alone vs. prior+attribute adapters vs. prior+text adapters; (2) relative-pose consistency (wrist-to-body and finger-to-wrist error under body-driven kinematics) with and without adapters; (3) the same suite under free attribute and free text control (not only body-conditioned completion); and (4) cross-dataset transfer of these scores. We will also report a “coupling dilution” delta (adapter-on minus prior-only) to show higher-layer adapters do not override the prior. These measurements will appear in a new table/figure and be referenced from the abstract and claims. This directly substantiates the frozen-prior advantage over end-to-end baselines.","revision_made":"yes","referee_comment":"The central claim that freezing the body-hand kinematic prior and injecting control only via lightweight semantically-layered adapters preserves mechanical body-hand coupling is load-bearing. Provide direct evidence—e.g., body-hand joint correlation or relative-pose consistency metrics with vs. without adapters, under attribute and text control, and in cross-dataset settings—that higher-layer adapters do not dilute or override the prior’s learned coordination."},{"response":"We agree the original wording was insufficiently precise. We will report exact labeled volumes (hours and frames) for every adapter setting (self-supervised attributes and weakly supervised text), and the matched training budgets (data scale, steps, compute) for all end-to-end conditioned baselines. We will add a controlled labeled-data scale ablation (e.g., 0.5 h / 1 h / few hours / full available labels) for both our adapters and the end-to-end baselines under identical optimization budgets, with a scale curve for kinematic and control metrics. This makes the efficiency and robustness claims assessable on equal footing and will replace the informal “few hours” phrasing with quantitative statements in abstract, method, and experiments.","revision_made":"yes","referee_comment":"The low-resource and “few hours of labeled data” claims are central. Report exact labeled data volumes (hours/frames) for each adapter setting, the corresponding training budgets and data scales for end-to-end baselines, and a controlled ablation that varies labeled-data scale. Without matched budgets and a scale curve, efficiency and robustness advantages cannot be fairly assessed."},{"response":"We agree. We will explicitly define the kinematic-plausibility suite: hand–body relative pose error, joint-limit violation rate, self-/body–hand interpenetration, temporal jerk and acceleration consistency, and the new body–hand coordination/correlation scores from Comment 1. We will state which of these drive the main claim and report them under (a) pure body-conditioned completion and (b) free attribute and free text control. Primary result tables will be revised so that reconstruction-style errors under body conditioning alone are no longer the sole support for the plausibility claim; coordination and constraint metrics under free semantic control will be first-class. The abstract and experimental narrative will be updated accordingly so that “improved kinematic plausibility” is unambiguously tied to these measurements.","revision_made":"yes","referee_comment":"Kinematic plausibility is multi-faceted. Clarify which metrics (hand-body relative pose error, joint-limit violations, interpenetration, temporal jerk/acceleration consistency, or analogous coordination scores) drive the “improved kinematic plausibility” claim, and whether they are evaluated under free semantic control as well as pure body-conditioned completion. If primary tables emphasize reconstruction-style errors under body conditioning alone, they do not fully address coupling preservation when adapters inject strong semantic signals."}],"tokens_in":1854,"tokens_out":1039,"duration_ms":15062,"standing_objections":[]},"desk_editor":{"model":"grok-4.5","letter":"Punchline: they learn a streaming body-hand kinematic prior from unlabeled motion, freeze it, then add lightweight layered adapters for semantic control. The bet is that kinematics come from scale and labels only need to steer, which is the right problem framing for scarce hand labels and real-time authoring.\n\nWhat is actually new is the packaging, not the pattern. Prior-then-adapt and LoRA-style adapters are established. Applying a streaming autoregressive body-hand prior with structured kinematic modeling, then injecting control at chosen kinematic levels for both self-supervised attributes and weakly supervised text with a few hours of labels, is a careful systems contribution for body-conditioned hand completion. The production angle—real-time inference and interactive authoring—is concrete and earned if the numbers hold. Circularity is low; this is empirical comparison work, not a fitted-constant story.\n\nThe soft spot that matters is the stress-test one, and it is real rather than pedantic. Freezing the prior and controlling only through light adapters assumes those adapters do not break the learned body-hand coupling. That is load-bearing. If higher semantic layers (especially text) dilute coordination under novel body dynamics or cross-dataset motion, the claimed gains in plausibility and robustness versus end-to-end baselines become regime-specific. Referees should demand coupling metrics with adapters on versus off, failure cases under strong control, and the usual ablations on adapter placement and rank. Everything else—architecture knobs, data scales—is ordinary systems detail. Novelty is moderate; significance is honest mid-subfield animation tooling, not field-reshaping.\n\nWho it is for: people building character animation pipelines, motion completion, and controllable generative motion. The design is clear and the problem is real. It deserves a serious referee, not a desk reject. I would send it out; the main job of review is to pressure whether the frozen prior still enforces mechanical coupling once control is applied.","headline":"Clean prior-then-adapt systems paper for hand completion; useful if the frozen prior still holds body-hand coupling once semantic adapters fire.","tokens_in":2936,"tokens_out":490,"would_cite":false,"duration_ms":20702,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.5","headline":"A frozen body-hand kinematic prior plus lightweight adapters completes hand motion that stays plausible and controllable","keywords":["hand motion completion","body-hand kinematic prior","autoregressive motion generation","lightweight adapters","semantic control","real-time animation","low-resource learning","motion synthesis"],"falsifier":"Train the same prior end-to-end with conditioning instead of freezing it and adding adapters; if kinematic error, cross-dataset robustness, and low-label text control match or beat the frozen-plus-adapters setup on the paper's benchmarks, the prior-first claim fails.","tokens_in":2979,"feed_emoji":"✋","tokens_out":766,"duration_ms":9980,"temperature":0.7,"pith_summary":"Synthesizing hand motion that matches full-body dynamics and semantic labels is hard because hands have many degrees of freedom and labeled data are scarce. This paper argues that the right order of operations is prior-first, condition-second: first learn a generic, streaming autoregressive body-hand kinematic prior from large unlabeled motion, capturing the intrinsic coordination between body and hands; then freeze that prior and inject control only through lightweight, semantically layered adapters. The adapters place conditioning at the right kinematic levels so that self-supervised attributes or a few hours of text labels can steer the hands without forcing the model to relearn kinematics. If the claim holds, animators and interactive systems get real-time, kinematically coherent hand completion that stays robust in low-resource and cross-dataset settings and supports interactive authoring—without the brittleness of end-to-end conditioned models trained on limited labeled data.","feed_headline":"Frozen body-hand prior plus adapters completes hands in real time","feed_subtitle":"Prior first, control second: better kinematic plausibility and text control from limited labels.","key_machinery":"The prior-first, condition-second pipeline: a frozen streaming autoregressive body-hand kinematic prior that maintains mechanical body-hand coupling, with semantically layered adapters that inject attribute or text control only at the right kinematic levels without relearning the prior.","core_discovery":"A streaming autoregressive body-hand kinematic prior learned from large unlabeled motion, kept frozen, plus lightweight semantically layered adapters that inject control at appropriate kinematic levels, yields more plausible, robust, and controllable body-conditioned hand motion completion than end-to-end conditioned baselines, especially under limited labels and across datasets, while remaining real-time.","pith_inferences":["The same prior-first pattern may transfer to other high-DoF, sparsely labeled body parts (face, feet) where kinematics dominate and semantics are scarce.","Interactive authoring could treat the adapters as live handles, letting artists dial attributes while the frozen prior keeps motion legal.","If unlabeled body-hand corpora grow, the prior could absorb new styles without touching existing control adapters.","Weak supervision budgets might shrink further if self-supervised attributes already cover most of the semantic axes needed for text control."],"forward_implications":["Hand completion stays kinematically coherent even when body motion is novel or out-of-distribution relative to labeled sets.","A few hours of text labels suffice for weakly supervised text-driven hand control without full joint retraining.","Real-time streaming inference supports interactive authoring in production animation pipelines.","Self-supervised attribute control can be layered at different kinematic levels without destroying body-hand coupling.","Cross-dataset transfer improves because the bulk of kinematic structure is learned once from unlabeled motion."],"fun_headline_variants":["Frozen body-hand prior then layered adapters complete hands real-time","Streaming autoregressive prior freezes body-hand coupling for control","Learn body-hand prior from unlabeled data then inject semantic adapters","Prior first control second: frozen kinematics plus adapters for hands","Real-time hand completion via frozen prior and semantically layered adapters"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"That freezing a generic body-hand kinematic prior and injecting control only through lightweight adapters at chosen kinematic levels is enough for semantic control, without relearning kinematics or breaking body-hand coupling when labeled data are scarce.","fun_headline_variants_meta":{"raw":{"variants":["Frozen body-hand prior then layered adapters complete hands real-time","Streaming autoregressive prior freezes body-hand coupling for control","Learn body-hand prior from unlabeled data then inject semantic adapters","Prior first control second: frozen kinematics plus adapters for hands","Real-time hand completion via frozen prior and semantically layered adapters"]},"model":"grok-4.5","cost_usd":0.014212,"raw_usage":{"total_tokens":2964,"prompt_tokens":797,"num_sources_used":0,"completion_tokens":69,"cost_in_usd_ticks":142120000,"prompt_tokens_details":{"text_tokens":797,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2098,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":797,"tokens_out":69,"duration_ms":19215,"temperature":1.0,"reasoning_tokens":2098,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-08T19:23:54.331084+00:00","model_set":{"reader":"grok-4.5"},"falsifier":"Train the same prior end-to-end with conditioning instead of freezing it and adding adapters; if kinematic error, cross-dataset robustness, and low-label text control match or beat the frozen-plus-adapters setup on the paper's benchmarks, the prior-first claim fails.","supporting_citations":[],"review_version":1}