Pith. sign in

REVIEW 4 major objections 5 minor 31 references

Diffusion-FS: Multimodal Free-Space Prediction via Diffusion for Autonomous Driving

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A diffusion model that denoises contour points, not pixel masks, predicts safe multimodal driving corridors directly from a single front-view camera image, trained without manual labels by projecting the vehicle's future trajectory into…

desk verdict New image-space corridor prediction task with a clever contour-diffusion architecture, but the evaluation labels come from the same heuristic that supervises training, so the safety claims need independent validation. read the letter →

arxiv 2507.18763 v1 pith:IVKN4LNE submitted 2025-07-24 cs.CV cs.RO

Keshav Gupta , Tejas S. Stanley , Pranjal Paul , Arun K. Singh , K. Madhava Krishna This is my paper
classification cs.CVcs.RO
keywords visualcorridorpredictionfree-spacediffusionmodelscontourdenoisingself-supervisedlearningmultimodalautonomousdrivingmonocularperception
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that drivable free-space corridor prediction can be done entirely in image space from a monocular front-view camera, without bird's-eye-view maps or dense manual annotations. The key move is self-supervised label generation: the ego vehicle's future trajectory is projected into the current image to define a free-space segment, which is clipped at the nearest overlapping obstacle. To model the inherently multimodal distribution of such corridors, the authors introduce ContourDiff, a diffusion model that denoises a set of contour points rather than a pixel mask. On CARLA and nuScenes, they report that ContourDiff outperforms YOLOv11 and SegDiff in IoU and obstacle overlap, and produces multiple plausible navigable corridors for the same scene.

What carries the argument

ContourDiff is the central object: a diffusion denoiser that takes a noisy contour, represented as N points in $R^{2}$, plus the input image, extracts image features at those point locations via bilinear sampling, concatenates positional embeddings and a sinusoidal timestep embedding, and refines the points through transformer self-attention layers and an MLP head. This contour-point parameterization carries the argument: it reduces the output space from H x W mask pixels to N x 2 points, imposes a structured, connected-segment prior, and makes every intermediate denoising step geometrically interpretable. The self-supervised label pipeline is the other load-bearing piece: future ego footprints from the driving log are projected into the camera frame, and the mask is clipped at the closest overlapping obstacle bounding box to guarantee the segment is collision-free.

What would settle it

On a held-out set of images, manually annotate all drivable corridors (not just the ego's future path). If, for scenes where the ego turned left but the right side is also clearly drivable, ContourDiff never produces a right-side corridor even with multiple samples, then the claim that it predicts the full multimodal navigable free space fails. Similarly, if a mask-based diffusion model plus a connected-component post-processing step matches or beats ContourDiff on the same IoU and obstacle-overlap metrics, the claimed advantage of contour-point denoising would be unsupported.

Watch

Extended reading notes

Core claim

The paper establishes visual corridor prediction as a pure perception task: given one front-view image, the model outputs an ordered set of contour points bounding a collision-free, navigable subset of the road, rather than classifying the entire drivable area. The central discovery is that a diffusion process operating on N x 2 contour points, with image features sampled at those points and mixed through transformer layers, learns the conditional distribution of free-space contours q(C|I) more effectively than binary-mask diffusion segmentation. The self-supervised data-generation procedure uses future ego footprints projected into the image, bounded by the closest obstacle, to create training pairs from raw driving logs. The result is a generative model that, at inference, denoises multiple noise samples into distinct safe corridors, capturing turn-left, turn-right, and straight options in the same scene.

Load-bearing premise

The training and evaluation labels are made by drawing the ego vehicle's future path into the image and cutting it off at the first obstacle that overlaps that path, which assumes the ego's future path is always inside the drivable free space and that clipping at the first overlap yields the complete safe corridor for the scene.

Editorial extensions

If this is right

  • Any camera-equipped vehicle with recorded trajectories can generate training data for free-space corridor prediction, removing the need for manually annotated segmentation datasets.
  • Because the model outputs multiple samples per image, a downstream planner can evaluate several navigable corridors (e.g., left turn, right turn, straight) before committing to one, matching the multimodal nature of driving decisions.
  • Denoising over contour points avoids the disconnected-mask artifacts that mask-based diffusion methods like SegDiff produce, yielding structured predictions that can be consumed directly as geometric constraints.
  • Class conditioning on high-level driving commands and obstacle guidance trade a small amount of IoU for higher prediction diversity and lower obstacle overlap, giving the user control over the generated multimodal outputs.
  • The approach transfers from a simulator (CARLA) to real-world data (nuScenes), suggesting it does not rely on simulation-specific cues.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Editorial inference: the self-supervised labels inherit a bias toward the exact paths the ego vehicle happened to take; regions that are visibly drivable but were never traversed in the training logs will be missing from the target distribution, so the model may under-predict valid corridors in novel traffic layouts.
  • Editorial inference: the method's ceiling depends on the recall of the obstacle detector that supplies bounding boxes, since any missed obstacle would cause the clipping step to produce labels that include that obstacle's region as free space.
  • Editorial inference: the noise-template initialization, borrowed from image-editing practice, suggests a general recipe for steering generative perception models toward desired modes, which could be transferred to other multimodal prediction tasks like trajectory forecasting or occupancy prediction.
  • Editorial inference: a natural testable extension is to feed the contour outputs directly into a planner in closed-loop driving on CARLA and measure whether the multimodal samples reduce collision rates compared to a single deterministic corridor predictor.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper proposes ContourDiff, a diffusion model that predicts multimodal free-space driving corridors directly in a front-view image by denoising a set of contour points conditioned on image features. Training labels are generated self-supervised by projecting the future ego-vehicle footprint into the camera image and clipping it at the closest detected obstacle. The model is evaluated on CARLA and nuScenes against YOLOv11 and SegDiff, with additional experiments on class conditioning, obstacle guidance, and noise-template initialization for multimodality. The main claim is that ContourDiff outperforms both baselines in IoU and obstacle overlap while producing diverse, safe corridor predictions.

Significance. If the results hold under independent ground truth, the paper would establish a useful image-space corridor prediction task that avoids BEV representations and dense annotations; the contour-point diffusion formulation is a clean and interpretable alternative to mask-based diffusion and could benefit downstream planning. The self-supervised label-generation idea is practical, and the choice of baselines (a non-generative segmenter and a mask-based diffusion segmenter) is appropriate. The strengths are the clearly stated task formulation and the structured contour representation, which is more parameter-efficient than mask diffusion. However, the current evaluation is largely circular because the same heuristic that creates training labels also defines the evaluation ground truth, the nuScenes off-road-overlap numbers contradict the safety narrative, and the multimodality metric does not validate that samples correspond to real route options. These issues make the significance conditional on additional validation.

major comments (4)
  1. [Section III-A and Table I] The pseudo-ground-truth generation is used both to create training labels and to compute all evaluation metrics, so the reported IoU, obstacle-overlap, and off-road-overlap numbers measure how well each model reconstructs the future ego-trajectory footprint clipped at the first overlapping obstacle, not whether the predicted regions are safe navigable corridors. Concretely, the clipping procedure only excludes the closest obstacle intersecting the projected footprint, so a prediction passing through a visible obstacle that does not intersect the recorded ego path would still receive a high IoU and low obstacle overlap; and the nuScenes off-road-overlap values in Table I (ContourDiff 0.21, YOLOv11 0.205) show that the recorded trajectories themselves frequently leave the drivable area, contradicting the assumption that the ego trajectory is always inside free space. Independent validation is needed, for example manually annotated free-space/corridor masks on a held-out subset or a closed-loop planner evaluation with collision rates.
  2. [Section IV-D and Table I] The statement that ContourDiff 'maintains low off-road overlap' is inconsistent with Table I: on nuScenes, ContourDiff's off-road overlap is 0.21, essentially equal to YOLOv11's 0.205 and roughly ten times SegDiff's 0.022. Because the safety claim of the paper rests on off-road avoidance, the authors should either explain why this high value is an artifact of the pseudo-label (for example, trajectories crossing sidewalks being labeled as free space) or report a corrected metric on data where the ego path is known to be inside the drivable area. If SegDiff's low value is caused by its frequent empty predictions, this should be reported together with prediction recall.
  3. [Section IV-F and Table III] The multimodality evidence is not tied to ground-truth route options. The Directional Deviation metric only measures the angular dispersion of six samples per image; a model producing arbitrary scattered contours would score well, and no experiment checks whether the samples correspond to the actual left/straight/right corridors or to the six high-level commands used for class conditioning. Furthermore, the noise-template method depends on the hyperparameters K and t, yet K is never assigned a value (only t=10 is reported), and using ground-truth contours to build initialization modes means the diversity results partly reflect the chosen templates rather than the diffusion model's learned multimodality. The authors should report K and evaluate multimodality against known route alternatives, for example by classifying the predicted direction per sample and measuring agreement with the available turning options.
  4. [Section IV-B and Table I] No error bars, multiple seeds, or statistical tests are reported, and the claimed improvements over baselines are small in some cells (for example, CARLA obstacle overlap 0.0200 versus YOLOv11 0.026). Given that training is stochastic and the evaluation uses a fixed split, the authors should provide results from at least three training seeds with mean and standard deviation for all metrics, or a paired test over images, to support the superiority claim.
minor comments (5)
  1. [Section III-B, Eq. (1)] The quantities alpha, gamma, and sigma are said to depend on the variance schedule but are never defined; providing the explicit schedule functions or a precise reference to the DDPM equations would make the method reproducible.
  2. [Section III-A, projection formula] The relation p' = K * R * [u, -h, v]^T appears to omit camera translation and assumes a particular ground-plane convention; please define the coordinate frames and extrinsic parameters fully.
  3. [Table I] The metrics are described as percentages but reported as fractions (0.02, 0.21, etc.); please state the units and normalization consistently.
  4. [Section IV-E] The explanation that the class-conditioned model's lower validation IoU is caused by a single ground-truth label is reasonable, but Table III would be more informative if it also reported the fraction of samples matching the commanded high-level behavior, for example by classifying the predicted contour direction.
  5. [Section IV-A] No information is given about which CARLA towns are used for evaluation, whether evaluation frames are temporally adjacent to training frames, or how many distinct episodes appear in the nuScenes validation split; please clarify to rule out overlap between training and evaluation logs.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the ego-trajectory label heuristic limits external validity but does not make the predictions equivalent to their inputs.

full rationale

The paper's central chain is: (1) Section III-A defines a target free-space mask S_t by projecting the future ego footprint into the image and clipping it at the nearest overlapping obstacle; (2) a diffusion model is trained to map the current image I to contours C_t via the objective in Eq. (2); (3) Table I compares predicted masks to the same S_t-derived masks. At inference, the future ego trajectory is not an input to the model, so the prediction is a nontrivial function of the image and noise; it is not the label-generation procedure itself. The shared heuristic between training labels and evaluation ground truth is a proxy-validity limitation, not a derivation equivalence: there is no equation in which the reported prediction reduces to the input by construction, and no fitted parameter is renamed as a prediction. The paper's statement that clipping 'guarantee[s] free-space' is an assumption baked into the self-supervised label, not a circular step in the model's derivation. The only self-citation, [16], appears in related-work context and is not load-bearing. Therefore, no circularity of the enumerated kinds is present; concerns about the safety interpretation of the pseudo-ground truth are better classified as external-validity or correctness risks rather than circularity.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced; ContourDiff is an architectural contribution. The central claim rests on the self-supervised label generation heuristic, whose assumptions are listed above, plus several unreported hyperparameters.

free parameters (5)
  • K (number of ground-truth contours averaged into noise templates)
    Used in Section IV-F to construct structured initializations; value not reported.
  • t (noise template diffusion timestep) = 10
    Set by hand in Section IV-F to balance structure and noise.
  • N (number of contour points) = 50
    Architecture choice in Section IV-B; defines the point-set representation.
  • Future trajectory horizon T for footprint generation
    Determines the length of the free-space segment; not specified in Section III-A.
  • Obstacle guidance gradient weight
    Controls strength of obstacle avoidance correction in Section IV-E2; not quantified.
assumptions (5)
  • standard math DDPM forward and reverse diffusion equations hold for contour point sets.
    Used in Section III-B to define training objective and sampling.
  • domain assumption The ego vehicle's future trajectory always lies inside drivable free space and is collision-free.
    Section III-A builds labels by projecting future footprints; invalid if the logged trajectory clips obstacles or goes off-road.
  • domain assumption Obstacle bounding boxes from an off-the-shelf detector accurately cover all obstacles in the image.
    Section III-A clips the free-space segment at the closest overlapping box; missed obstacles would create unsafe labels.
  • domain assumption The ground is locally planar so BEV points project to the image via p' = K R [u, -h, v].
    Section III-A uses this projection; violations on slopes or curbs distort labels.
  • ad hoc to paper Averaging K ground-truth contours per high-level command and adding noise produces valid initialization modes.
    Section IV-F uses noise templates to improve multimodality; no evidence these templates cover all valid corridors.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Diffusion-FS: Multimodal Free-Space Prediction via Diffusion for Autonomous Driving." pith.science (2026). https://pith.science/paper/IVKN4LNE

@misc{pith2026250718763,
  author       = {Pith},
  title        = {Pith review of: Diffusion-FS: Multimodal Free-Space Prediction via Diffusion for Autonomous Driving},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IVKN4LNE}},
  note         = {Machine review of arXiv:2507.18763}
}
read the original abstract

Drivable Free-space prediction is a fundamental and crucial problem in autonomous driving. Recent works have addressed the problem by representing the entire non-obstacle road regions as the free-space. In contrast our aim is to estimate the driving corridors that are a navigable subset of the entire road region. Unfortunately, existing corridor estimation methods directly assume a BEV-centric representation, which is hard to obtain. In contrast, we frame drivable free-space corridor prediction as a pure image perception task, using only monocular camera input. However such a formulation poses several challenges as one doesn't have the corresponding data for such free-space corridor segments in the image. Consequently, we develop a novel self-supervised approach for free-space sample generation by leveraging future ego trajectories and front-view camera images, making the process of visual corridor estimation dependent on the ego trajectory. We then employ a diffusion process to model the distribution of such segments in the image. However, the existing binary mask-based representation for a segment poses many limitations. Therefore, we introduce ContourDiff, a specialized diffusion-based architecture that denoises over contour points rather than relying on binary mask representations, enabling structured and interpretable free-space predictions. We evaluate our approach qualitatively and quantitatively on both nuScenes and CARLA, demonstrating its effectiveness in accurately predicting safe multimodal navigable corridors in the image.

Figures

Figures reproduced from arXiv: 2507.18763 by the authors.

Figure 1
Figure 1. Left: A dataset of raw driving logs containing image and ego trajectory pairs. Our self-supervised method processes such an unannotated dataset to generate free-space segments essential for autonomous driving. Right: Examples of multimodal free-space segments generated by our diffusion model on CARLA. At inference, our model denoises a fixed number of 6 noise samples into free-space segments. We showcase predictions… view at source ↗
Figure 3
Figure 3. Free-space Contour Creation. Left : We show the transformations between the ego vehicle’s frame at time t + k and the frame at time t. Right: We show the process of creation of the free-space sample for an image. The top row presents the BEV map in the local frame of the ego vehicle at time t, while the middle and bottom rows show the corresponding frontal camera images and an alternative top-down view, illustrating… view at source ↗
Figure 4
Figure 4. Conditional Probability Distribution of Free-space Con￾tours given an image. We show an example of an intersection where the distribution of free-space contours is likely to be bimodal, as there is possibility of free-space at both the left and the right turn. The training data provides enough evidence to approximate this distribution, as in many driving logs covering a similar scenario, the ego vehicle must have tr… view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: The architecture of the proposed ContourDiff. The image I and the initial noisy contour C Tmax are passed as input to the model. Note that C t is visualized on top of image, and is not part of the image. The output of the model is the denoised contour C 0 which is obta…
Figure 6
Figure 6. Figure 6: Training Samples : We show different samples generated on CARLA and nuScenes, on applying the methodology described in Section III various driving scenarios such as straight roads, inter￾section turns, lane changes, and lane following. Our collected dataset comprises 8…
Figure 7
Figure 7. Figure 7: Top Row: CARLA – Comparison of YOLOv11, SegDiff, and our proposed ContourDiff at an intersection. The non-generative baseline YOLOv11 struggles to predict the free-space segment. We present six samples from both SegDiff and ContourDiff, demonstrating that ContourDiff, …
Figure 8
Figure 8. Figure 8: Effect of denoising from a set of predefined noise templates vs random noise template in the base model. Left: Multimodality tends to increase as different modes are explored with different noise template initializations. Right: With random noise as initialization, the…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

31 extracted references · 28 canonical work pages

  1. [1]

    ViNT: A foundation model for visual navigation,

    D. Shah et al., “ViNT: A foundation model for visual navigation,” in CoRL, 2023

    work page 2023

  2. [2]

    Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,

    P. Wu et al., “Trajectory-guided control prediction for end-to-end autonomous driving: A simple yet strong baseline,” in NeurIPS, 2022

    work page 2022

  3. [3]

    Neat: Neural attention fields for end-to-end autonomous driving,

    K. Chitta, A. Prakash, and A. Geiger, “Neat: Neural attention fields for end-to-end autonomous driving,” in ICCV, 2021

    work page 2021

  4. [4]

    Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly un- projecting to 3d,

    J. Philion and S. Fidler, “Lift, splat, shoot: Encoding images from arbitrary camera rigs by implicitly un- projecting to 3d,” in ECCV, 2020

    work page 2020

  5. [5]

    End-to-end interpretable neural mo- tion planner,

    W. Zeng et al. , “End-to-end interpretable neural mo- tion planner,” in ICCV, 2019

    work page 2019

  6. [6]

    Planning-oriented autonomous driving,

    Y . Hu et al., “Planning-oriented autonomous driving,” in ICCV, 2023

    work page 2023

  7. [7]

    TwinLiteNet+: An Enhanced Multi-Task Segmentation Model for Autonomous Driving

    Q.-H. Che et al., “Twinlitenetplus: A real-time multi- task segmentation model for autonomous driving,” arXiv preprint arXiv:2403.16958 , 2024

    work page Pith review arXiv 2024

  8. [8]

    Drivable area detection using deep learning models for autonomous driving,

    D. Qiao and F. Zulkernine, “Drivable area detection using deep learning models for autonomous driving,” in IEEE International Conference on Big Data , 2021

    work page 2021

Show all 31 references
  1. [9]

    Drivespace: Towards context- aware drivable area detection,

    C. Eising et al. , “Drivespace: Towards context- aware drivable area detection,” Electronic Imaging , vol. 2019, pp. 42–1, Jan. 2019

    2019

  2. [10]

    Pyramid scene parsing network,

    H. Zhao et al. , “Pyramid scene parsing network,” in CVPR, 2017

    2017

  3. [11]

    Real-time capable deci- sion making for autonomous driving using reachable sets,

    N. Kochdumper and S. Bak, “Real-time capable deci- sion making for autonomous driving using reachable sets,” in ICRA, 2024

    2024

  4. [12]

    Sampling-based op- timal trajectory generation for autonomous vehicles using reachable sets,

    G. W ¨ursching and M. Althoff, “Sampling-based op- timal trajectory generation for autonomous vehicles using reachable sets,” in IEEE International Intelligent Transportation Systems Conference (ITSC) , 2021

    2021

  5. [13]

    Using reach- able sets for trajectory planning of automated ve- hicles,

    S. Manzinger, C. Pek, and M. Althoff, “Using reach- able sets for trajectory planning of automated ve- hicles,” IEEE Transactions on Intelligent Vehicles , vol. PP, pp. 1–1, Aug. 2020

    2020

  6. [14]

    Towards responsibility-sensitive safety of automated vehicles with reachable set analysis,

    P. F. Orzechowski, K. Li, and M. Lauer, “Towards responsibility-sensitive safety of automated vehicles with reachable set analysis,” in IEEE International Conference on Connected Vehicles and Expo , 2019

    2019

  7. [15]

    Trimodal navigable region seg- mentation model: Grounding navigation instructions in urban areas,

    N. Hosomi et al. , “Trimodal navigable region seg- mentation model: Grounding navigation instructions in urban areas,” RAL, vol. 9, no. 5, pp. 4162–4169, 2024

    2024

  8. [16]

    Lego-drive: Language-enhanced goal- oriented closed-loop end-to-end autonomous driving,

    P. Paul et al., “Lego-drive: Language-enhanced goal- oriented closed-loop end-to-end autonomous driving,” in IROS, 2024

    2024

  9. [17]

    Yolov11: An overview of the key architectural enhancements,

    R. Khanam and M. Hussain, “Yolov11: An overview of the key architectural enhancements,” arXiv preprint arXiv:2410.17725, 2024

    2024 arXiv

  10. [18]

    Faster r-cnn: Towards real-time object detection with region proposal networks,

    S. Ren et al., “Faster r-cnn: Towards real-time object detection with region proposal networks,” NeurIPS, 2015

    2015

  11. [19]

    You only look once: Unified, real- time object detection,

    J. Redmon et al., “You only look once: Unified, real- time object detection,” in CVPR, 2016

    2016

  12. [20]

    Segdiff: Image segmentation with diffusion probabilistic models,

    T. Amit et al. , “Segdiff: Image segmentation with diffusion probabilistic models,” arXiv preprint arXiv:2112.00390, 2021

    2021 arXiv

  13. [21]

    Diffumask: Synthesizing images with pixel-level annotations for semantic segmentation us- ing diffusion models,

    W. Wu et al. , “Diffumask: Synthesizing images with pixel-level annotations for semantic segmentation us- ing diffusion models,” in ICCV, 2023

    2023

  14. [22]

    Label-efficient semantic seg- mentation with diffusion models,

    D. Baranchuk et al. , “Label-efficient semantic seg- mentation with diffusion models,” in ICLR, 2022

    2022

  15. [23]

    High-resolution image synthesis with latent diffusion models,

    R. Rombach et al. , “High-resolution image synthesis with latent diffusion models,” in CVPR, 2022

    2022

  16. [24]

    Medsegdiff: Medical image segmenta- tion with diffusion probabilistic model,

    J. Wu et al. , “Medsegdiff: Medical image segmenta- tion with diffusion probabilistic model,” in Medical Imaging with Deep Learning , PMLR, 2024

    2024

  17. [25]

    Ambiguous medical image seg- mentation using diffusion models,

    A. Rahman et al. , “Ambiguous medical image seg- mentation using diffusion models,” in CVPR, 2023

    2023

  18. [26]

    Scaling open-vocabulary object detection,

    M. Minderer, A. Gritsenko, and N. Houlsby, “Scaling open-vocabulary object detection,” in NeurIPS, 2023

    2023

  19. [27]

    Topological structural analysis of digitized binary images by border following,

    S. Suzuki and K. be, “Topological structural analysis of digitized binary images by border following,” Com- puter Vision, Graphics, and Image Processing, vol. 30, no. 1, pp. 32–46, 1985

    1985

  20. [28]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in NeurIPS, 2020

    2020

  21. [29]

    Nuscenes: A multimodal dataset for autonomous driving,

    H. Caesar et al., “Nuscenes: A multimodal dataset for autonomous driving,” in CVPR, 2020

    2020

  22. [30]

    Learning from all vehi- cles,

    D. Chen and P. Kr ¨ahenb¨uhl, “Learning from all vehi- cles,” in CVPR, 2022

    2022

  23. [31]

    SDEdit: Guided image synthesis and editing with stochastic differential equations,

    C. Meng et al. , “SDEdit: Guided image synthesis and editing with stochastic differential equations,” in ICLR, 2022

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.