{"id":"6ad50321-68b5-409d-9b72-68e219074eaf","arxiv_id":"2610.08780","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":4,"one_line_summary":"A calibration pipeline turns DROID into a robot-grounded metric-depth dataset, and joint RGB-depth training of a diffusion world model improves RGB prediction while providing metric depth.","lead":"DROID-3D recalibrates 71,100 robot episodes from the DROID dataset into a metric-depth corpus grounded in each robot's own body frame, and DepthWorld is a diffusion world model that predicts RGB and depth together by tiling both modalities into a single latent grid. Jointly training on the two signals improves RGB generation by 1.48 dB PSNR over an RGB-only model at equal budget while also producing depth, a property policy evaluation and planning depend on.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek/deepseek-v4-flash","headline":"Headline RGB gain is not yet attributable to depth: RGB+D uses six tiled latents (3 RGB + 3 depth) versus three for the RGB-only baseline, doubling spatial tokens and per-step attention compute; no token-matched RGB-only control exists, so the +1.48 dB may be capacity rather than depth-supervision.","rationale":"I read the paper's center of gravity as the claim that depth supervision improves RGB prediction while yielding depth, because that is what b157 calls 'the central empirical finding' and what the architecture section is built around. The reader's weakest assumption (S2M2 self-reproduction) is real and is explicitly acknowledged at b343 and Appendix C.8; it constrains the absolute-depth claim but does not touch the RGB gain. The token-count confound is different: it directly threatens the controlled RGB comparison and is not acknowledged anywhere. The paper has genuine independent support—ChArUco and REALM calibration validation, a retrained equal-split baseline, and a plausible architectural story via Marigold—so I would not reject the paper. But the headline needs one additional control before the attribution is proven. I therefore keep the verdict CONDITIONAL, with the added condition that the authors supply a token-matched, no-depth baseline. My concern is not about novelty or style; it is about whether the single controlled experiment in the paper actually isolates the variable it names.","tokens_in":30350,"tokens_out":10116,"duration_ms":135206,"concrete_test":"Retrain an RGB-only control on the 72×80 grid, filling the three depth-tile slots with (i) per-step i.i.d. Gaussian noise, (ii) a constant gray image, and (iii) a duplicate of the same view's RGB tile, keeping step count, batch size, and optimizer identical to RGB+D. Report external-view PSNR on the same 256 held-out trajectories and FLOPs/step for each variant. If any no-depth control moves PSNR materially from 22.63 toward 24.09, the headline is a capacity/token-count effect; only if all remain near 22.63 can 'depth supervision improves RGB' be accepted.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central empirical claim (b157) is the +1.48 dB PSNR gain of RGB+D over an 'identical' RGB-only baseline, but the comparison is not confound-free. Section 4.1 describes RGB+D as a joint latent of 72×80 made of six tiles (RGB and depth for each of three views); the RGB-only baseline has only three RGB tiles, so the latent grid is half as wide in tokens and the U-Net's per-step compute (notably attention) is correspondingly smaller. Appendix D.3 calls the baseline retrained 'on the same set, split, architecture, and training budget', but 'budget' is never defined as steps vs FLOPs; if it means steps, the depth model receives substantially more compute per step. No ablation equalizes token count while withholding depth information: Table A2 compares joint RGB-depth architectures (dual-branch, channel expansion, tiling), not an RGB-only model with the depth tiles replaced by noise, a constant, or duplicated RGB. The claimed attribution of the RGB gain to depth supervision therefore rests on an untested assumption. The paper's Limitations (b153) do not mention this confound.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper proposes DROID-3D, a recalibration of the DROID dataset into a robot-frame 3D corpus using learned stereo (S2M2) depth plus a joint factor graph that pools episodes per robot to recover shared hand-eye and joint-offset parameters, and DepthWorld, a Stable Video Diffusion world model that jointly predicts multi-view RGB and depth by spatially tiling VAE latents and optionally supervising a DPT point-map head in the robot base frame. The central claims are that DROID-3D provides metric depth and recalibrated extrinsics with large improvements over a PointWorld-style per-scene baseline, and that depth supervision through spatial tiling improves RGB prediction by +1.48 dB PSNR over an RGB-only baseline at equal training budget, while also yielding depth for geometric reasoning. The paper includes extensive ablations, held-out trajectory evaluation, ChArUco and REALM pose validation, and comparisons with PointWorld and an adapted TesserAct baseline.","tokens_in":30640,"tokens_out":4291,"duration_ms":57672,"significance":"If the claims hold, the paper is a substantial practical contribution: it upgrades a widely used real-robot manipulation dataset with dense metric depth and robot-grounded multi-view calibration, and it shows a way to inject 3D supervision into a pretrained video-diffusion world model without destructive architectural modification. The ChArUco/REALM validation of the calibration pipeline (4.9 mm / 0.39 deg median extrinsics error) is a strong piece of evidence, and the paper is unusually candid about caveats such as the VAE round-trip evaluation floor and the held-out-trajectory (not cross-embodiment) scope. The main risk is that the headline RGB improvement is not yet attributable to depth supervision because the RGB+D model also has twice the latent token count of the RGB-only baseline; this needs a token-matched control before the central claim can be accepted at face value.","major_comments":[{"comment":"The headline +1.48 dB PSNR gain is confounded by model capacity. DepthWorld encodes six tiles (RGB and depth for each of three views) into a 72×80 joint latent, while the RGB-only baseline uses three RGB tiles in a 72×40 grid; this doubles the spatial token count and, with attention in the U-Net, roughly doubles per-step compute. Appendix D.3 calls the baseline a controlled comparison with the 'same set, split, architecture, and training budget,' but the training budget is never defined as steps versus FLOPs, and the two models do not receive the same compute per step at equal step count. No control replaces the three depth tiles with noise, a constant, or duplicated RGB while keeping the token grid at 72×80. Until such a control is run, the +1.48 dB gain cannot be attributed to depth supervision rather than to the additional learned capacity, and the Limitations paragraph (b153) does not acknowledge this confound.","section":"§4.1 and Appendix D.3 (Table 2, sentence b157)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The metric-depth claim is only sparsely grounded and is otherwise self-referential. All DROID-3D depth, for both training and evaluation, comes from the learned S2M2 disparity network; the depth metrics in Table 2 and Appendix C are computed against the VAE round-trip of the same S2M2 depth, so they measure self-reproduction of the training signal, exactly as the authors note in b343. Independent metric grounding is provided for five in-lab ChArUco capture sets and 50 REALM scenes, but 71,100 episodes remain validated only against S2M2 itself. The conclusion that DepthWorld produces 'accurate metric depth for downstream geometric reasoning' therefore holds only under the untested assumption that S2M2 depth is metrically unbiased on the unvalidated DROID scenes. I would like to see either a larger independent depth-validation set or a clearer separation between the calibrated-grounding claim and the S2M2-reproduction claim in the abstract and Section 5.","section":"Appendix C.8 and Appendix A.1 (Table 2, sentence b343)","authors_acknowledge":"The depth metrics are not: each model is scored against the depth source it was trained on, so they mea- sure how faithfully a model reproduces its own supervision rather than its absolute metric accuracy. 11 Table A2:Ablation: Spatial latent tiling vs. other integration architectures, at∼70k training steps under the autoregressive rollout protocol (Appendix D.3).","authors_acknowledge_p":0.74},{"comment":"The comparison with PointWorld is qualified in the text but the table presentation makes it easy to over-read. In Table A5, PointWorld is scored against raw ground-truth depth while DepthWorld is scored against VAE-round-tripped ground truth, and Appendix D.4 notes that the VAE round trip alone displaces the ground truth by about 45 mm on moving points. This 'own-ceiling' protocol is disclosed and argued carefully, but the headline numbers (e.g., whole-scene Chamfer 8→9 mm for DepthWorld vs 8→18 mm for PointWorld) mix two different targets. Since the off-diagonal comparisons are said to reverse the winner, the summary in Section 5 should state explicitly that the comparative ranking is conditional on each model being scored above its own representational floor, and the main text should not present Table A5 as a direct accuracy comparison.","section":"§5 and Appendix D.4 (Table A5)","authors_acknowledge":"Second, DepthWorld predicts depth in the latent space of the SVD V AE, whose encode–decode round trip alone has an error floor that PointWorld’s point representation does not have; how we account for this is described below.","authors_acknowledge_p":0.83},{"comment":"Several reported differences between RGB+D and RGB+D PM-DPT are very small (24.09 vs 24.11 dB PSNR; external AbsRel 0.0765 vs 0.0782), and no variance across seeds or confidence intervals is reported anywhere. Given that the PM-DPT contribution is advertised as a refinement, single-run differences of this size are not interpretable. Reporting at least two seeds or bootstrap intervals for the main Table 2 metrics would make the qualitative claims actionable.","section":"§5, Table 2","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"There is a duplicated word in 'kinematically decoupled from from the robot’s own coordinate frame'; please remove the second 'from.'","section":"§1, Introduction","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The caption says DepthWorld 'capitalizes on the depth inputs,' but the inference protocol described in Appendix C.7 initializes the history buffer from the first frame and the model is not given ground-truth depth at rollout. Please clarify whether depth is an input at inference or only a training target, and adjust the caption wording accordingly.","section":"Figure 1 caption","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The phrase 'specially on wrist metrics' should be 'especially on wrist metrics.' Also, the depth-quality columns in Table 2 are evaluated against the VAE round-trip of the S2M2 depth; this caveat appears in Appendix C.8 but should be repeated in the Table 2 caption or Section 5 protocol for readers who do not read the appendix.","section":"Table 2 caption and §5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The text uses unresolved placeholder citations for CoTracker3, DBSCAN, and CogVideoX-5b-I2V ('[?]'). These need to be replaced with proper references before publication.","section":"Appendix D.4 and D.5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The statement 'Spatial tiling performs better: it improves every RGB metric on both views' refers to Table A2, but that table compares joint RGB-depth architectures and does not include a token-matched RGB-only baseline; please add a sentence making this limitation explicit in the appendix as well as in the main text.","section":"Appendix C.2","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"Colleague — the real product here is the calibration pipeline, and it's good. Pooling all episodes of one physical robot into a single factor graph to separate shared kinematic parameters (hand-eye, joint offsets) from per-scene extrinsics is a genuine scaling insight: the Schur-complement reduction to a 13x13 global solve is clean, and the results back it up, 4.9 mm / 0.39 deg against ChArUco ground truth versus 72 mm / 3.0 deg for the PointWorld-style per-scene baseline. The calibration table (23x EE, 13x WE, 8x RD, +37 Mask IoU) is convincing, and the own-ceiling protocol in the PointWorld comparison is honest, careful evaluation work. DROID-3D, if released, is a real asset for the subfield.\n\nThe headline RGB claim is where I part company with the framing. The stress-test note is right: RGB-only runs three latent tiles per timestep, RGB+D runs six. Same nominal budget is not the same compute — attention scales with token count — and the paper never defines 'budget' as steps or FLOPs, and never runs the control that would settle it: an RGB-only model with the depth tiles replaced by noise, a constant, or duplicated RGB. The +1.48 dB is a real measured difference, but attributing it to depth supervision rather than extra capacity is untested, and the Limitations section does not mention the confound. One nuance in the paper's favor: the depth-source ablation (Table A3) shows bad depth actively hurts RGB (-2 dB), so the depth signal is doing something. But the S2M2-vs-ZED-NEURAL row, same token count, near-identical RGB at 24.07 vs 24.26 dB, is equally consistent with the gain being mostly tokens. Needs a matched-token control before I'd trust the causal story.\n\nOther soft spots, in proportion. No code or data shipped, and DROID-3D itself unreleased, so nothing is independently reproducible. All depth metrics are scored against the S2M2 depth used in training — disclosed in C.8 — but the 'accurate metric depth at 71,000 episodes' claim rests on only five ChArUco sets and fifty REALM scenes of external grounding. The unresolved [?] citations for CoTracker3, CogVideoX-5b-I2V, and DBSCAN in Appendices D.4-D.5 break replication of the baseline protocols. The dual-branch comparison is admitted to be not 1:1.\n\nVerdict: send to a serious referee, yes. The calibration contribution stands on its own even if the RGB-attribution story collapses. I would ask for a token-matched RGB-only control and code or data before accepting.","headline":"The calibration contribution is real and well-evidenced, but the headline RGB gain is confounded: RGB+D runs twice the latent tokens per step, and the paper never runs the token-matched control that would clean the attribution.","tokens_in":31130,"tokens_out":3594,"duration_ms":22131},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"Jointly predicting depth alongside RGB — via a spatial-latent tiling scheme that leaves a pretrained Stable Video Diffusion backbone untouched — improves RGB video itself by 1.48 dB PSNR over an identical RGB-only model, while also…","keywords":["3D world models","robot manipulation","video diffusion","metric depth calibration","factor graph optimization","spatial latent tiling","DROID dataset"],"falsifier":"Run the full S2M2-and-calibration pipeline on a much larger independent sample of DROID scenes and compare arm-surface depth against the URDF render at the recovered extrinsics, or sweep a ChArUco board through many more labs than the five validated: if the independently measured median depth error on the full corpus is well above the reported ~14 mm robot-depth error, the claim that DROID-3D is metrically grounded at scale collapses. Separately, the RGB-gain claim is settled by a control training run in which the depth targets are permuted or replaced with matched noise: if the +1.48 dB PSNR gain persists, it is not specifically depth that helps.","tokens_in":31517,"feed_emoji":"🤖","tokens_out":6635,"duration_ms":42552,"temperature":0.7,"pith_summary":"The paper tries to establish that a video world model for robot manipulation can produce rollouts that compose into a single coherent 3D scene, not just visually plausible frames. To get there it builds two things: DROID-3D, a recalibration of the DROID dataset that supplies dense metric depth and robot-grounded camera poses for over 70,000 episodes, and DepthWorld, a video-diffusion world model that predicts multi-view RGB and depth together. The central empirical claim is that adding depth as a joint prediction target improves RGB prediction itself by +1.48 dB PSNR at equal training budget, so geometric supervision is not a cost to appearance prediction but a benefit. If correct, this opens a route to world models whose rollouts can be read off for policy evaluation, improvement, and planning.","feed_headline":"Depth-trained world model beats RGB-only on RGB","feed_subtitle":"DROID-3D recalibration plus spatial latent tiling yields coherent 3D rollouts with metric depth.","key_machinery":"Spatial latent tiling is the architectural device that carries the argument: each view's RGB and depth are encoded as separate images by the untouched SVD VAE and placed side-by-side in a 72x80 latent grid, so the single pretrained U-Net denoises both modalities jointly with only the spatial position embeddings extended. This is what allows depth supervision to be added without destructive re-initialization of the pretrained video backbone. The companion mechanism is the joint factor graph, an SE(3)-parameterized robust loss over all episodes of a robot whose Schur-complement solve reduces to a 13x13 global problem per iteration, cleanly separating per-scene camera corrections from shared kinematic parameters. A DPT point-map head initialized from VGGT lifts the predicted RGB-depth latents into robot base coordinates and supervises cross-view agreement directly in 3D.","core_discovery":"The core claim is that 3D-consistent world-model rollouts are attainable without sacrificing the pretrained video prior. On the data side, the paper shows that a multi-view stereo teleoperation collection with a known robot kinematic model can be lifted into supervision-quality 3D: S2M2 learned stereo disparity supplies per-view metric depth, and a joint factor graph pooling all episodes of one physical robot separates per-scene extrinsic noise from shared kinematic errors (hand-eye correction and joint offset), grounding the rig in the robot's base frame. On the modeling side, DepthWorld encodes each view's RGB and depth independently with the unmodified SVD variational autoencoder and tiles them side-by-side in a wider latent grid, so a single pretrained U-Net denoises both modalities together. The paper reports that this joint target beats an identical RGB-only baseline by +1.48 dB PSNR on external views and +1.02 dB on wrist views while producing metric depth, with a lightweight DPT point-map head supervised in the robot base frame further sharpening wrist-view depth.","pith_inferences":["The +1.48 dB RGB gain is scored against real RGB pixels, so it would survive even if S2M2 depth were metrically biased; it requires only that depth is an information-bearing auxiliary signal. A cleaner control than the paper runs would corrupt or permute the depth targets and check the gain disappears.","If tiling works for depth, it likely generalizes to any modality the pretrained VAE can encode — normals, segmentation, optical flow — giving a generic recipe for joint-modality world models.","The tracked-point and Chamfer comparisons with PointWorld hint that depth-image world models drift less on static scenes than point-cloud world models; whether that holds over longer horizons and harder scenes is left open."],"forward_implications":["World models for manipulation can be trained so their rollouts carry internally consistent geometry, making occlusion and contact reasoning possible where RGB-only models fail.","Supervising a second modality need not trade off against the first: depth supervision improved the RGB signal, so geometric training targets can act as a free gain.","The calibration approach transfers to any multi-view stereo teleoperation corpus with a known URDF, so other large manipulation datasets could be lifted into 3D without new capture hardware.","Because tiling exploits the VAE's latent capacity rather than re-initializing weights, the pattern could transfer to other pretrained video diffusion backbones.","The recalibrated extrinsics have a measurable downstream effect: point-map supervision trained against DROID-3D poses rather than factory poses improves wrist-view depth accuracy (AbsRel 0.233 to 0.220).","The paper reports that the calibration is far more robust than per-scene refinement, with no analogous per-scene failure mode to the baseline's roughly 18% non-convergence."],"supporting_citations":[{"why":"The DROID dataset: supplies the ~71k raw teleoperation episodes across 28 robots and 13 labs that DROID-3D recalibrates and DepthWorld trains on.","marker":"[18]"},{"why":"S2M2: the learned stereo disparity network whose output is the metric depth ground truth for both the calibration pipeline and every DepthWorld depth metric.","marker":"[35]"},{"why":"Stable Video Diffusion: the pretrained video diffusion backbone and variational autoencoder that DepthWorld preserves via spatial latent tiling.","marker":"[23]"},{"why":"Marigold: the observation that an image-diffusion VAE encodes and decodes depth faithfully, which motivates encoding depth as a tiled image.","marker":"[24]"},{"why":"Ctrl-World: the action-conditioned backbone DepthWorld starts from and the RGB-only baseline that defines the equal-budget comparison.","marker":"[1]"},{"why":"PointWorld: the per-scene calibration baseline that the joint factor graph must beat, and the 3D world model benchmarked in Section 5.","marker":"[34]"},{"why":"LoMA-G: the render-and-match tool that supplies 2D-3D correspondences anchoring the external cameras to the robot's URDF frame.","marker":"[37]"},{"why":"DUSt3R: supplies the confidence-weighted log-space robust point-map loss used by the auxiliary DPT head.","marker":"[6]"},{"why":"VGGT: initializes the DPT point-map head and supplies the fusion trunk reused in the robot-frame auxiliary supervision.","marker":"[8]"}],"fun_headline_variants":["Depth supervision boosts RGB in 3D world model","DepthWorld: joint RGB-depth beats RGB-only by 1.48 dB","Depth-trained world model yields coherent 3D rollouts","DROID-3D recalibration powers DepthWorld's metric depth"],"cache_read_input_tokens":0,"weakest_assumption_plain":"Every depth number in the paper — the calibration targets that define DROID-3D and the ground truth that all DepthWorld depth accuracy is scored against — comes from the learned S2M2 stereo disparity, which is checked against real metric ground truth on only five in-lab ChArUco capture sets and 50 REALM scenes out of 71,100 total.","fun_headline_variants_meta":{"raw":{"variants":["Depth supervision boosts RGB in 3D world model","DepthWorld: joint RGB-depth beats RGB-only by 1.48 dB","Depth-trained world model yields coherent 3D rollouts","DROID-3D recalibration powers DepthWorld's metric depth"]},"model":"deepseek/deepseek-v4-flash","effort":"low","cost_usd":0.000149,"raw_usage":{"cost":0.000149268,"is_byok":false,"cost_details":{"upstream_inference_cost":0.000149268,"upstream_inference_prompt_cost":0.000112588,"upstream_inference_completions_cost":3.668e-05},"total_tokens":1140,"prompt_tokens":1009,"completion_tokens":131,"prompt_tokens_details":{"audio_tokens":0,"video_tokens":0,"cached_tokens":256,"cache_write_tokens":0},"completion_tokens_details":{"audio_tokens":0,"image_tokens":0,"reasoning_tokens":56}},"tokens_in":1009,"tokens_out":131,"cost_source":"openrouter_usage","duration_ms":1956,"image_pages":[],"temperature":1.0,"billed_attempts":1,"reasoning_tokens":56,"transport_provider":"openrouter","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-10-07T03:13:04.217468+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":"Run the full S2M2-and-calibration pipeline on a much larger independent sample of DROID scenes and compare arm-surface depth against the URDF render at the recovered extrinsics, or sweep a ChArUco board through many more labs than the five validated: if the independently measured median depth error on the full corpus is well above the reported ~14 mm robot-depth error, the claim that DROID-3D is metrically grounded at scale collapses. Separately, the RGB-gain claim is settled by a control training run in which the depth targets are permuted or replaced with matched noise: if the +1.48 dB PSNR gain persists, it is not specifically depth that helps.","supporting_citations":[{"cited_title":"DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset","cited_arxiv_id":"2403.12945","evidence_quote":"The DROID dataset: supplies the ~71k raw teleoperation episodes across 28 robots and 13 labs that DROID-3D recalibrates and DepthWorld trains on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"S2M2: the learned stereo disparity network whose output is the metric depth ground truth for both the calibration pipeline and every DepthWorld depth metric."},{"cited_title":"Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets","cited_arxiv_id":"2311.15127","evidence_quote":"Stable Video Diffusion: the pretrained video diffusion backbone and variational autoencoder that DepthWorld preserves via spatial latent tiling."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Marigold: the observation that an image-diffusion VAE encodes and decodes depth faithfully, which motivates encoding depth as a tiled image."},{"cited_title":"Ctrl-World: A Controllable Generative World Model for Robot Manipulation","cited_arxiv_id":"2510.10125","evidence_quote":"Ctrl-World: the action-conditioned backbone DepthWorld starts from and the RGB-only baseline that defines the equal-budget comparison."},{"cited_title":"PointWorld: Scaling 3D World Models for In-The-Wild Robotic Manipulation","cited_arxiv_id":null,"evidence_quote":"PointWorld: the per-scene calibration baseline that the joint factor graph must beat, and the 3D world model benchmarked in Section 5."},{"cited_title":"LoMa: Local Feature Matching Revisited","cited_arxiv_id":"2604.04931","evidence_quote":"LoMA-G: the render-and-match tool that supplies 2D-3D correspondences anchoring the external cameras to the robot's URDF frame."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"DUSt3R: supplies the confidence-weighted log-space robust point-map loss used by the auxiliary DPT head."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"VGGT: initializes the DPT point-map head and supplies the fusion trunk reused in the robot-frame auxiliary supervision."}],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":null},"url":"https://pith.science/accuracy"},"review_version":1}