{"id":"220b61ed-775f-4424-95ca-0c4a4f0283df","arxiv_id":"2610.36840","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Auxiliary future 3D motion prediction improves the robustness and few-shot performance of World Action Models for robot manipulation.","lead":"Trace-WAM adds future 3D motion prediction as an extra training objective for robot learning, helping the model focus on how objects and the robot move during manipulation. The method achieves higher success rates on standard benchmarks and real-world tasks, especially when training data is limited or the environment changes.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"The auxiliary 3D motion objective is confounded with increased capacity and multi-task training; no control task equates model size and data to isolate the motion-specific contribution.","rationale":"The reader flagged annotation‑quality as the weakest assumption, which is reasonable but does not threaten the simulated benchmarks where annotations are perfect. A more incisive concern is whether the motion objective's content (temporal point‑correspondence + depth change) is causally responsible for the gains, or whether any auxiliary task with similar capacity and training would produce comparable improvements. The paper lacks this control. The proposed test (depth‑map prediction as a non‑motion auxiliary task) directly isolates the motion‑specific contribution. If the control matches Trace‑WAM, the paper's central rationale weakens, though the empirical finding of improved performance remains. Therefore, a conditional verdict is appropriate: the work is strong, but the core interpretation needs additional empirical support before full acceptance.","tokens_in":16252,"tokens_out":3219,"duration_ms":34636,"concrete_test":"Replace the motion expert with an expert that predicts future depth maps (projected from the video decoder's latent) using the same token count, training schedule, and loss structure (e.g., smoothness penalty). Train on LIBERO, evaluate zero‑shot on LIBERO‑Plus. If this depth‑prediction baseline reaches ≥84.5% average success, the benefit is not specific to 3D motion; if it falls substantially, the motion‑specific claim is supported.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The paper attributes gains to the specific 3D motion prediction objective, but the motion expert adds parameters and FLOPs, and multi-task training can regularize and improve generalization regardless of task content. Ablations remove internal components (e.g., depth, B-spline, static penalty) but never swap the motion objective for a different auxiliary prediction task with comparable capacity and training schedule. Without such a control, observed improvements could stem from any structured auxiliary supervision, not the temporal point-correspondence and depth-change representation argued as essential. This is load‑bearing because the central claim (Conclusion, [b181]) asserts that future 3D motion prediction focuses representation learning on manipulation‑relevant dynamics; if a different auxiliary task yields similar gains, the argument for motion‑specific representation is unsupported. The reader's concern about annotation quality is valid but partially mitigated by simulation benchmarks; the lack of a task‑agnostic control is a more fundamental threat to causal interpretation.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper proposes Trace-WAM, which augments World Action Models (WAMs) with future 3D motion prediction as an auxiliary objective. A three-branch Mixture-of-Transformer (Tri-MoT) architecture jointly learns video prediction, motion prediction, and action generation. Motion supervision is applied only during training; inference uses current observations. Experiments on RoboCasa-GR1, LIBERO, and LIBERO-Plus show improved in-distribution and out-of-distribution performance over baselines, with particularly clear gains under limited demonstrations and distribution shifts. Real-world experiments on a Piper robot also report higher success rates.","tokens_in":16603,"tokens_out":1975,"duration_ms":18046,"significance":"If the reported gains are causally attributable to the 3D motion prediction objective, the work demonstrates a practical way to focus representation learning on manipulation-relevant dynamics, which could generalise across tasks and domains. The benchmarks (LIBERO, RoboCasa-GR1, LIBERO-Plus) are well-established, and the real-world validation strengthens the claim. The paper is clearly written and the ablation studies are structured. However, the central interpretation is weakened by the absence of a control for the auxiliary-task confound.","major_comments":[{"comment":"The paper attributes performance gains to the specific 3D motion prediction objective (Conclusion, [b181]: 'auxiliary future 3D motion prediction to focus video–action representation learning on the scene dynamics most relevant to manipulation'). However, the ablations in Table 4 compare variants of the motion supervision (e.g., w/o depth, w/o B-spline) but never replace the motion objective with a different auxiliary prediction task (e.g., predicting future depth maps, optical flow, or a dummy regression) that would match the motion expert in model capacity and training schedule. Without such a control, the observed improvements could arise from any structured auxiliary supervision or simply from increased model capacity, rather than the specific temporal-point-correspondence and depth-change representation. This is a load-bearing issue for the central claim.","section":"§4.3 (Ablation) and §5 (Conclusion)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The motion expert adds a substantial number of parameters (separate Transformer with hidden dimension 1,024) compared to the Fast-WAM baseline, which has no such expert. The paper does not control for increased capacity (e.g., by enlarging Fast-WAM's video or action expert to match total parameters, or by adding a parameter-matched module with a different objective). The improved performance may partly reflect additional capacity rather than the motion prediction task. A controlled comparison is needed to isolate the role of the auxiliary objective.","section":"§4 (Experimental Setup) and §3.2 (Motion Representation)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"For real-world demonstrations, 3D motion labels are obtained from the TraceForge pipeline, which estimates depth, camera motion, and point tracks from video. The paper states that 'the supplementary material describes the complete annotation and validation procedure,' but the main text provides no quantitative error analysis or cross-validation of these estimates. Given that the real-world results are used to support the method's generality, the accuracy of the annotation pipeline is load-bearing for that part of the evaluation. At minimum, a summary of validation metrics (e.g., point-tracking reprojection error, depth accuracy) should be reported in the main paper.","section":"§3.2 (last paragraph) and §4.4 (Real-world Experiments)","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The title on page 1 reads 'Trace-W AM' with a space in 'W AM'. This appears inconsistently throughout (e.g., 'WAM' elsewhere). Please unify.","section":"Title","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Figure 1 caption uses 'Trace-W AM' with a space in the second occurrence; please ensure consistent formatting.","section":"Figure 1 caption","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Table 5 reports motion prediction accuracy on LIBERO-Plus. The caption says '140 rollout episodes' but the text does not specify how rollouts were selected or whether they cover all perturbation types. Clarify.","section":"Table 5 footnote","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The sentence 'This difference motivates a complementary target that emphasizes interaction dynamics while retaining the rich visual context learned from future RGB' is somewhat vague. Consider rephrasing for clarity.","section":"§1 Introduction","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The description of B-spline fitting (Eq. 3) mentions scaling 'each displacement channel using training-set statistics' but does not specify which statistics (mean/std or min/max). Please clarify.","section":"§3.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The motion expert uses a convolution with kernel and stride 2×2 to map 20×20 to 100 tokens. The text says 'each token contains four complete future trajectories.' Provide a brief justification for this design choice or reference.","section":"§3.3","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Trace-WAM does what good incremental papers do: it identifies a plausible gap in an existing paradigm (WAMs learn video prediction that wastes capacity on background), designs a clean auxiliary objective (3D point trajectories with depth, B-spline parameterization, static-motion penalty), and evaluates it thoroughly. The results hold up: consistent gains on RoboCasa, LIBERO, LIBERO-Plus, and a real robot, particularly in few-shot and OOD settings where the margins are meaningful. The ablations in Table 4 are well-structured and support each design choice individually. The paper is clearly written, the architecture is sensible, and the inference path is unchanged from Fast-WAM.\n\nThe main soft spot is the one flagged by the stress-test: there is no control task that matches model capacity and training schedule but replaces the motion objective with a different structured prediction (e.g., predicting depth maps, segmenting static objects, or even a synthetic target). Without that, the core attribution—that 3D motion prediction specifically focuses representations on manipulation-relevant dynamics—rests on weaker ground. Any auxiliary multi-task objective that adds parameters and regularizes may produce similar gains here, especially given that the LIBERO in-distribution results are near ceiling. The paper's ablations remove components of the motion objective but never replace it with a different auxiliary task of comparable complexity. This doesn't invalidate the results, but it does mean the central mechanistic claim in [b181] and [b185] is only partly supported.\n\nA secondary concern: the TraceForge pipeline for real-world motion annotations is described but not error-analyzed or cross-validated (Section 3.2 last paragraph). Given that depth estimation and point tracking are noisy in the wild, a brief analysis of annotation quality would bolster confidence in the real-robot results.\n\nThese are not fatal flaws. The paper makes a real practical contribution, the experiments are reproducible in principle, and the gains in OOD and few-shot settings are worth having. A reviewer should ask for a control auxiliary-task experiment (even a simple one) and a note on annotation accuracy, but the paper is publishable as is. For someone working on video-conditioned robot policies, this is a useful read.","headline":"Trace-WAM is a solid, well-engineered extension to World Action Models; the gains are real and the experiments are thorough, but the paper overclaims specificity to 3D motion since it lacks a multi-task control to rule out generic auxiliary supervision benefits.","tokens_in":17139,"tokens_out":1782,"duration_ms":19216},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"Trace-WAM augments robot policy learning with auxiliary 3D motion prediction, improving performance under limited data and distribution shifts.","keywords":["World Action Models","robot manipulation","3D motion prediction","point tracking","auxiliary supervision","flow matching","generalization","video prediction"],"falsifier":"If a controlled experiment replaced the predicted motion targets with random noise (while keeping the architecture and training unchanged) and achieved similar task success on LIBERO-Plus, the specific motion content would not be the cause of improvement. Alternatively, if the motion-prediction accuracy (ADE and MAE) did not correlate with task success across different seeds, the claim that motion supervision drives the gain would be weakened.","tokens_in":17766,"feed_emoji":"🤖","tokens_out":3618,"duration_ms":30765,"temperature":0.7,"pith_summary":"Trace-WAM is a method for training robot policies that jointly learns to predict future video, future 3D motion, and actions. The key idea is that predicting how points move in 3D (including depth) over time forces the model to represent the interaction dynamics most relevant to manipulation. This auxiliary objective is used only during training; at inference, the policy acts from current observations alone. Experiments on three benchmarks show consistent improvements over baselines, especially with few demonstrations and under visual distribution shifts. Real-world tests on a Piper robot confirm the gains.","feed_headline":"Motion prediction lifts robot policy success 18 points","feed_subtitle":"Future-3D-motion supervision improves World Action Models without extra inference cost, shown on three benchmarks and a real robot.","key_machinery":"The central mechanism is the future 3D motion prediction auxiliary objective: a grid of 400 anchor points in the current image, each tracked over 32 future steps with their image-plane coordinates and depth changes, represented compactly via cubic B-spline coefficients. The Tri-MoT architecture has three Transformer experts—video, motion, and action—with structured attention masks that allow motion and future-video tokens to attend to current video but prevent current video and action tokens from attending to motion or future video. This ensures the motion supervision shapes representations without adding inference-time computation.","core_discovery":"The paper establishes that auxiliary future 3D motion prediction—supervising temporal point correspondence with image-plane displacement and depth change—provides a powerful learning signal for World Action Models. This approach, implemented via a three-branch Mixture-of-Transformer (Tri-MoT) architecture, improves both in-distribution performance and out-of-distribution generalization. The motion objective is decoupled from the action path at inference, preserving the efficiency of a single forward pass. Results on RoboCasa-GR1, LIBERO, and LIBERO-Plus show consistent gains, with the largest improvements in few-shot and OOD settings, such as +18.8 points over Fast-WAM in few-shot RoboCasa and +18.3 points on LIBERO-Plus overall.","pith_inferences":["The motion prediction objective might be extended to predict future contact points or force profiles, enabling policies that reason about physical interaction more explicitly.","The reliance on the TraceForge pipeline for real-world motion annotations means that annotation quality is a potential bottleneck; a rigorous error analysis of that pipeline would clarify the sensitivity of the method.","The grid-based motion representation assumes a fixed camera viewpoint for the main camera; a multi-view or dynamic camera setting might require modification."],"forward_implications":["If the approach generalizes, robot policies trained with fewer demonstrations can achieve higher success rates, reducing the data collection burden for new tasks.","The auxiliary motion objective could be applied to other visuomotor architectures beyond the specific World Action Model framework used here.","Because the motion representation includes depth, the method may be particularly robust to background and lighting changes that do not affect 3D motion.","The B-spline parameterization of trajectories provides a compact target that improves prediction accuracy and task success over raw per-step displacements."],"supporting_citations":[{"why":"Provides the Fast-WAM baseline and training codebase that Trace-WAM extends.","marker":"[48]"},{"why":"Introduces 4D tracking for world action models, a related approach using motion distillation.","marker":"[42]"},{"why":"Defines the grid-based motion representation used for anchor placement.","marker":"[39]"},{"why":"Introduces the B-spline representation for motion trajectories.","marker":"[20]"},{"why":"Baseline vision‑language‑action model evaluated for comparison.","marker":"[3]"},{"why":"Provides a strong baseline (DIAL) on RoboCasa-GR1 for comparison.","marker":"[7]"},{"why":"Defines the LIBERO benchmark used for in‑distribution evaluation.","marker":"[27]"},{"why":"Defines the RoboCasa-GR1 benchmark used for few‑shot and OOD evaluation.","marker":"[32]"}],"fun_headline_variants":["Trace-WAM: 3D motion prediction lifts policy 18 points","Auxiliary future 3D motion boosts robot manipulation 18 pts","Robot policies gain 18 points with future 3D motion prediction","Efficient robot learning via future 3D motion prediction 18-pt gain"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The training requires accurate 3D motion annotations (point tracks with depth) for every demonstration, and for real-world data these are estimated by the TraceForge pipeline without a quantitative error analysis in the paper.","fun_headline_variants_meta":{"raw":{"variants":["Trace-WAM: 3D motion prediction lifts policy 18 points","Auxiliary future 3D motion boosts robot manipulation 18 pts","Robot policies gain 18 points with future 3D motion prediction","Efficient robot learning via future 3D motion prediction 18-pt gain"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000107,"raw_usage":{"cost":0.00010725708,"is_byok":false,"cost_details":{"upstream_inference_cost":0.00010725708,"upstream_inference_prompt_cost":4.568172e-05,"upstream_inference_completions_cost":6.157536e-05},"total_tokens":1530,"prompt_tokens":914,"completion_tokens":616,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":0,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":535}},"tokens_in":914,"tokens_out":616,"cost_source":"openrouter_usage","duration_ms":6860,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":535,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-28T13:35:25.700188+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium","input_kind":null,"input_chars":null,"input_total_chars":null,"passes_missing":null,"published_standing":null,"published_confidence":null,"objection_audit":null,"run_record":null,"basis":{"kind":"not_recorded","source":null,"chars_given":null,"chars_total":null},"receipt":null,"falsifier":"If a controlled experiment replaced the predicted motion targets with random noise (while keeping the architecture and training unchanged) and achieved similar task success on LIBERO-Plus, the specific motion content would not be the cause of improvement. Alternatively, if the motion-prediction accuracy (ADE and MAE) did not correlate with task success across different seeds, the claim that motion supervision drives the gain would be weakened.","supporting_citations":[{"cited_title":"Fast-WAM: Do World Action Models Need Test-time Future Imagination?","cited_arxiv_id":"2603.16666","evidence_quote":"Provides the Fast-WAM baseline and training codebase that Trace-WAM extends."},{"cited_title":"4D-WAM: Infusing Spatiotemporal Awareness into World Action Models through Trajectory Fields","cited_arxiv_id":"2608.08023","evidence_quote":"Introduces 4D tracking for world action models, a related approach using motion distillation."},{"cited_title":"Lamp: Learning vision-language-action policy with 3d scene flow as la tent m otion p rior","cited_arxiv_id":null,"evidence_quote":"Defines the grid-based motion representation used for anchor placement."},{"cited_title":"$\\mu_0$: A Scalable 3D Interaction-Trace World Model","cited_arxiv_id":"2606.13769","evidence_quote":"Introduces the B-spline representation for motion trajectories."},{"cited_title":"$\\pi_0$: A Vision-Language-Action Flow Model for General Robot Control","cited_arxiv_id":"2410.24164","evidence_quote":"Baseline vision‑language‑action model evaluated for comparison."},{"cited_title":"DIAL: Decoupling Intent and Action via Latent World Modeling for End-to-End VLA","cited_arxiv_id":"2603.29844","evidence_quote":"Provides a strong baseline (DIAL) on RoboCasa-GR1 for comparison."},{"cited_title":"Libero: Benchmarking knowl- edge transfer for lifelong robot learning.Advances in Neural Information Processing Systems, 36:44776–44791, 2023","cited_arxiv_id":null,"evidence_quote":"Defines the LIBERO benchmark used for in‑distribution evaluation."},{"cited_title":"RoboCasa: Large-Scale Simulation of Everyday Tasks for Generalist Robots","cited_arxiv_id":"2406.02523","evidence_quote":"Defines the RoboCasa-GR1 benchmark used for few‑shot and OOD evaluation."}],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}