EditFlow3D: Automated Local Editing of 3D Assets with Trajectory Preservation

Buyu Li; Chuang Wang; Haitao Zhou; Jiahe Song; Qian Yu; Rui Nie; Sheng Wang

REVIEW 3 major objections 3 minor 175 references

EditFlow3D: Automated Local Editing of 3D Assets with Trajectory Preservation

T0 review · 3 major / 3 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 EditFlow3D claims that a masked difference between source- and guidance-conditioned flow velocities can drive precise local 3D edits, while a step-wise trajectory-preservation loss keeps non-target regions faithful to the source asset.

desk verdict A solid training-free 3D local-editing pipeline with a genuinely new preservation loss and a broad new benchmark, but the headline number rests on a benchmark that filters out the hard cases—so treat the automated-pipeline claim as conditional until they release code and unfiltered results. read the letter →

arxiv 2608.03179 v1 pith:7GRW2ZTV submitted 2026-08-04 cs.CV

Rui Nie , Chuang Wang , Haitao Zhou , Jiahe Song , Buyu Li , Sheng Wang , Qian Yu This is my paper

classification cs.CV

keywords 3DasseteditinglocaldifferentialflowguidancetrajectorypreservationrectifiedVLM-drivencontrolmaskrefinementtraining-free

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

EditFlow3D aims to make local 3D asset editing as simple as typing an instruction, without retraining or per-asset optimization. It claims that the edit can be concentrated in one region by taking the velocity difference a rectified-flow generative model predicts under the source view versus an edited guidance image, masking that difference to the target voxels, and then softly pulling non-target features back toward the encoded source state at every generation step. The paper argues that this combination—differential flow guidance plus trajectory preservation guidance—resolves the central tension of local editing: changing exactly what was asked while leaving the rest visibly untouched. If the claim holds, a user could replace, delete, add, or reshape parts of a 3D asset in under a minute from a text prompt and an automatically generated mask. The paper also contributes a benchmark of 100 assets and 200 editing cases across five operation types to test local editing more broadly.

What carries the argument

The central object is the differential velocity $\Delta v_t = \mathbb{E}_\epsilon[ v_\theta(z_{\text{tgt},t}, t, c_{\text{edit}}) - v_\theta(z_{\text{src},t}, t, c_{\text{src}}) ]$, computed by the flow estimator under target versus source visual conditions. The editable state is updated as $z_{\text{edit}} \leftarrow z_{\text{edit}} + \gamma \Delta t\, S_{\text{edit}} \odot \Delta v_t$, so the mask $S_{\text{edit}}$ turns a globally conditioned velocity difference into a local edit. The complementary mechanism is the preservation loss $\mathcal{L}_{\text{pres}} = \| S_{\text{keep}} \odot (\hat{x}_{0,t} - x_{\text{src}}) \|_2^2$, whose gradient softly corrects non-target features toward the

What would settle it

Run EditFlow3D on a case where the guidance image has been deliberately edited outside the target region, for example an instruction to replace a chair seat while the image editor also changes a chair leg's color, and give the pipeline a mask covering only the seat. If the final asset keeps the leg unchanged, the claim that differential velocity acts globally would be wrong; if the leg drifts, the paper's assumed dependence on guidance-image fidelity is confirmed.

Watch

Extended reading notes

Core claim

EditFlow3D asserts that a local edit to a 3D asset can be driven by the difference between two velocity predictions of a pretrained rectified-flow generative model—one conditioned on the source view, one on an edited guidance image—provided the differential velocity is masked to the target region. It further asserts that preserving non-target content does not require replacing or injecting source features; instead, a step-wise soft constraint that pulls predicted clean states in the keep-region toward the encoded source state suffices. The paper demonstrates this combination, called mask-guided differential flow guidance (DFG) and trajectory preservation guidance (TPG), in the native sparse

Load-bearing premise

The pipeline assumes that the source view chosen by the vision-language model and the image editor's guidance picture fully and correctly describe the intended local change; because the flow difference is conditioned on the whole images, any unintended alteration in the guidance picture enters the edit direction before the 3D mask is applied.

Editorial extensions

If this is right

Local editing becomes a one-shot, training-free operation in the native 3D representation space: no per-asset inversion, no optimization, and no fine-tuning of the generator.
A single edited 2D view plus a 3D mask is enough visual and spatial control; cross-view consistency is inherited from the 3D generator rather than stitched together from multiple 2D edits.
Non-target preservation does not require hard feature replacement, so edited and retained regions remain coherent across the boundary instead of showing seams or fragmentation.
The same flow-editing core works across two backbones of the structured-latent generator family, indicating the mechanism transfers with the flow formulation rather than with one specific network.
Automated vision-language controls plus mask refinement make the method applicable to relational instructions such as “the left pillar” or “the front smokestack,” not only simple part names.

Reading between the lines

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

If the differential-velocity recipe is truly backbone-agnostic, any conditional rectified-flow model with a structured latent space could be retrofitted with this editing loop, making the paper's mechanism a general template rather than a one-off method.
Because the velocity difference is conditioned on full images, the guidance image must be trustworthy outside the edit region; an obvious stress test is to feed a guidance image with an unintended change elsewhere and measure how much non-target drift appears.
The open loop between mask construction and final result invites a closed-loop variant: re-render the output and ask the same vision-language model whether the requested change is complete and nothing else moved, then regenerate if not.
Used repeatedly, the method could generate paired source–edited assets with verified untouched regions, which could serve as training data for future feedforward editors.

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.

Desk Editor's Note

read the letter

The core idea here is sound and worth knowing: EditFlow3D combines mask-guided differential flow with a new step-wise trajectory preservation loss (TPG) that softly pulls non-target features toward the source state, avoiding the boundary artifacts of hard feature injection. That is a real, plausible improvement over prior flow-editing work, and the ablations back it up—removing TPG noticeably hurts preservation metrics, and replacing it with direct injection causes visible fragmentation. The multi-model mask refinement is also a reasonable engineering contribution, and the new EditFlow-Bench covers broader edit categories than Edit3D-Bench. They also report results on Edit3D-Bench with oracle masks, which gives a somewhat cleaner read on the editing mechanism itself.

The main soft spot is the evaluation of the automated pipeline. EditFlow-Bench is explicitly filtered: guidance images are discarded and regenerated unless they preserve camera and non-target identity with no severe artifacts, and masks are manually inspected before use. That means the reported numbers estimate performance conditional on near-perfect upstream controls, not the unconditional performance of the 'automated framework' the abstract advertises. The paper's own limitation section concedes that view-selection, grounding, and segmentation errors propagate to the final edit. So the comparative advantage over baselines may persist on unfiltered inputs, but the current evidence does not establish it. There is also no code or benchmark data released, which makes the numbers unverifiable.

Two smaller flags: the preservation strength lambda is tuned on a single example ("Delete the fire in the stove"), which is a thin basis for a global hyperparameter; and the user study, though run at reasonable scale, inherits the benchmark filtering. None of this is fatal—the method is still plausible and the ablation design is honest—but it does temper the headline.

Who gets value from this? Anyone working on 3D asset editing or on evaluation practices for agentic editing pipelines. The benchmark filtering discussion is actually a useful cautionary example for the community. I would send it to peer review with a clear request for code/data release, a report of the rejection/regeneration rate, and ideally an unfiltered subset where upstream failures are shown rather than excluded. It deserves referee time, but the revision needs to close the gap between the claim and the evidence.

Referee Report

3 major / 3 minor

Summary. EditFlow3D proposes a training-free, inversion-free framework for local 3D asset editing. It first uses a VLM (Qwen3-VL-8B) to select a source view from multi-view renders, generates a guidance image with FLUX.1 Kontext, and constructs a 3D editing mask by fusing P3-SAM, PartField, and S2AM3D segmentations with graph-based boundary refinement. The editing stage applies Differential Flow Guidance (DFG): the difference between source- and target-conditioned velocity predictions is masked and used to update the editable latent in the native TRELLIS representation space. Trajectory Preservation Guidance (TPG) then softly constrains non-target regions toward the source state via a gradient of a preservation loss at each feature-generation step. The authors introduce EditFlow-Bench, a 200-case benchmark spanning addition, deletion, replacement, geometry modification, and appearance modification, and report quantitative metrics, ablations, and a user study showing improvements over Nano3D, PartFlow, VoxHammer, and Vinedresser3D on both EditFlow-Bench and Edit3D-Bench.

Significance. If the results hold, the paper makes a useful contribution: it offers a modular, training-free recipe for local 3D editing in a native latent representation, with a two-stage control pipeline that combines mask guidance with trajectory preservation. The component ablations support the roles of DFG and TPG, the method is evaluated on two TRELLIS-family backbones, and the external Edit3D-Bench results are a strong positive signal. The proposed benchmark is also a practical resource, though its construction protocol substantially limits what the reported numbers can establish about end-to-end automation. The main weakness is that the benchmark filters out cases where the upstream controls (guidance image or 3D mask) are imperfect, while the paper's central claim is about an automated pipeline; the current evidence is conditional on near-perfect upstream controls.

major comments (3)

[Datasets and App. B.2] EditFlow-Bench accepts a case only when the FLUX guidance image 'preserves the camera and non-target identity' and contains no severe artifact; failures are discarded and regenerated. Masks are also manually inspected and failed cases regenerated (App. B.3). Since the headline claim is an automated end-to-end pipeline, the quantitative results are conditional on near-perfect upstream controls. The paper's own limitation statement (App. D.2) concedes that incorrect view selection, ambiguous part grounding, and segmentation errors propagate to the final edit. Please report pass/rejection rates, evaluate on an unfiltered run, or explicitly qualify the claim so that the numbers are not over-generalized.
[App. C.3 and Table 2] The TPG strength lambda=12 is selected using a single case ('Delete the fire in the stove'). Table 2 shows that removing TPG degrades PSNR from 36.53 to 27.95 and FID from 43.51 to 67.84, so the preservation mechanism is a major contributor to the reported improvements. With the hyperparameter tuned on one example, the reported advantage may be optimistic. Provide a sensitivity analysis across the five edit types or justify the chosen value with a principled criterion.
[Eq. (6), Eq. (7), Eq. (10)] The differential velocity is computed from globally conditioned velocity fields; the 3D mask is applied only to the update in Eq. (7). Any unintended change in the FLUX guidance image therefore enters the editing direction in every voxel, and TPG's loss (Eq. (9)) is a soft correction on one predicted clean state. The benchmark's acceptance filter (App. B.2) removes exactly the cases where this failure mode would occur, so the experiments do not demonstrate that TPG can compensate for guidance-image leakage. I suggest adding an unfiltered or adversarially corrupted guidance-image stress test.

minor comments (3)

[General] No code or benchmark data are released. Since the paper introduces a new benchmark, releasing the benchmark cases, masks, and evaluation scripts would be important for reproducibility and for verifying the pass/fail protocol.
[App. A.2] The implementation parameters (tau_P=0.6, Kmax=8, lambda_G=1, eta=4, alpha_comp=0.005, alpha_hole=0.02, ring counts) are fixed without sensitivity analysis. A short discussion of their stability would strengthen the mask-refinement component.
[Evaluation Metrics] The computation of 'Sim.' (CLIP-based similarity on the edited region) and 'DINO-I' is not fully specified in the main text. Please clarify the exact rendering and masking procedures used to compute these metrics.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the editing mechanism is self-contained and benchmark evidence is independent of the method's construction.

full rationale

The paper's core derivation chain is not circular. Differential Flow Guidance (DFG, Eqs. 4-7) is explicitly adopted from the external FlowEdit principle, with the 3D mask applied as a spatial constraint; Trajectory Preservation Guidance (TPG, Eqs. 8-10) is a soft regularizer that pushes non-target predicted clean states toward the encoded source, not a fitted parameter or a renamed version of the reported preservation metrics. No parameter is fitted to the benchmark outcomes, and the central claims are evaluated on the public Edit3D-Bench in addition to the authors' EditFlow-Bench, with all baselines given the same source asset, guidance image, and mask protocol. The manual inspection and regeneration of EditFlow-Bench cases (Appendix B.2) and the conceded dependence on upstream controls (Appendix D.2) are external-validity and selection-bias concerns, not circularity: they affect how broadly the reported numbers generalize, but they do not make the method's predictions equivalent to its inputs by construction. There is no load-bearing self-citation chain, no imported uniqueness theorem, and no known result renamed as a new mechanism. A non-finding is therefore appropriate.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The method relies on several pretrained models and hand-picked hyperparameters but introduces no new physical entities or fitted constants that define the output. The main fitted quantity is lambda, tuned on one example.

free parameters (5)

TPG preservation strength lambda = 12
Chosen by visual inspection on a single case ('Delete the fire in the stove') among {0,6,12,20}; used fixed in all experiments.
DFG editing strength gamma = not stated
The update in Eq. (7) includes gamma*dt but the paper never reports the default value used in the experiments.
Purity threshold tau_P = 0.6
Hyperparameter for the candidate pool in mask refinement (Eq. 2); no sensitivity analysis is provided.
Mask expansion rings = 1 ring for deletion/appearance, 4 rings for replacement/geometry
Operation-specific expansion chosen by hand; directly controls the spatial extent of the edit and thus the preservation metrics.
Graph refinement hyperparameters lambda_G, eta, alpha_comp, alpha_hole = 1, 4, 0.005, 0.02
Set without ablation for the normal-aware boundary refinement in Appendix A.2.

assumptions (4)

domain assumption Rectified flow linear interpolation (1-t)x + t epsilon and the ODE z' = v(z,t) are valid for TRELLIS.
Eq. (4) and (8) rely on the pretrained flow model being a rectified flow with linear paths; the paper does not verify this against the specific TRELLIS version.
domain assumption The difference of velocity fields between source and target conditions defines a valid editing direction (FlowEdit principle).
Used in Eq. (6); adopted from Kulikov et al. without independent validation in the 3D latent space.
domain assumption The preservation loss gradient keeps non-target features on the source trajectory without destabilizing the target edit.
Eq. (9)-(10); validated only empirically on their benchmark, and the sensitivity to lambda shows it can fail at lambda=20.
domain assumption P3-SAM, PartField, and S2AM3D segmentations can be mapped to a common surface and their consensus approximates the true target region.
Appendix A.2; this is the basis of the 3D mask, and the paper acknowledges segmentation errors as a limitation.

how reviews work

0 comments

Cite this review

Pith. "Pith review of EditFlow3D: Automated Local Editing of 3D Assets with Trajectory Preservation." pith.science (2026). https://pith.science/paper/7GRW2ZTV

@misc{pith2026260803179,
  author       = {Pith},
  title        = {Pith review of: EditFlow3D: Automated Local Editing of 3D Assets with Trajectory Preservation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7GRW2ZTV}},
  note         = {Machine review of arXiv:2608.03179}
}

read the original abstract

Controllable local editing of 3D assets requires precise target localization and appropriate visual guidance. However, existing methods lack a simple yet accurate way to obtain 3D masks and struggle to achieve the desired edit while faithfully preserving the structure and appearance of non-target regions. To address these challenges, we present EditFlow3D, a training-free framework for local 3D editing. Given a source asset and an edit instruction, a VLM-driven workflow interprets the editing intent and automatically constructs a visual guidance image and a refined 3D editing mask, enabling localized editing in the native representation space of a pretrained 3D generative model. Specifically, mask-guided differential flow focuses the edit on the target region, while step-wise trajectory preservation maintains consistency between non-target regions and the source asset without directly replacing intermediate features. Since the existing Edit3D-Bench covers only a limited range of local editing categories, we further introduce EditFlow-Bench as a complementary benchmark encompassing a broader variety of structural and appearance edits, and evaluate EditFlow3D on both benchmarks. Quantitative results, qualitative comparisons, and a user study demonstrate that EditFlow3D achieves more accurate target-region editing and better preserves non-target regions than existing 3D editing methods.

Figures

Figures reproduced from arXiv: 2608.03179 by the authors.

**Figure 1.** We present EditFlow3D, a training-free and inversion-free framework for precise part-level editing of high-fidelity 3D assets. EditFlow3D supports localized structural edits such as part replacement and removal while preserving the geometry and appearance of unedited regions. and refines a 3D editing mask, thereby automatically producing accurate visual and spatial controls. In the second stage, Differential Flow G… view at source ↗

**Figure 3.** Qualitative comparison on representative local 3D part-editing cases. All editing methods use the same source asset and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 5.** Comparison between trajectory preservation guid [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Qualitative ablation on EditFlow3D. Each variant [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

175 extracted references · 17 canonical work pages

[1]

Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)
[2]

Classification Problem Solving

Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence
[3]

, title =

Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

1980
[4]

New Ways to Make Microcircuits Smaller---Duplicate Entry

Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science
[5]

Clancey and Glenn Rennels , abstract =

Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

doi:10.1016/s0020-7373(84)80003-6 1984
[6]

and Rennels, Glenn R

Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies
[7]

Poligon: A System for Parallel Problem Solving

Rice, James. Poligon: A System for Parallel Problem Solving
[8]

Transfer of Rule-Based Expertise through a Tutorial Dialogue

Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue

Show all 175 references

[9]

The Engineering of Qualitative Models

Clancey, William J. The Engineering of Qualitative Models
[10]

2023 , eprint=

Attention Is All You Need , author=. 2023 , eprint=

2023
[11]

Pluto: The 'Other' Red Planet

NASA. Pluto: The 'Other' Red Planet
[12]

arXiv preprint arXiv:2210.02747 , year=

Flow matching for generative modeling , author=. arXiv preprint arXiv:2210.02747 , year=

arXiv
[13]

Computer Vision -- ECCV 2022 , year =

2022
[14]

arXiv preprint arXiv:2512.14692 , year=

Native and compact structured latents for 3d generation , author=. arXiv preprint arXiv:2512.14692 , year=

arXiv
[15]

arXiv preprint arXiv:2507.06165 , year=

OmniPart: Part-Aware 3D Generation with Semantic Decoupling and Structural Cohesion , author=. arXiv preprint arXiv:2507.06165 , year=

arXiv
[16]

arXiv preprint arXiv:2207.12598 , year=

Classifier-free diffusion guidance , author=. arXiv preprint arXiv:2207.12598 , year=

arXiv
[17]

Communications of the ACM , volume=

Nerf: Representing scenes as neural radiance fields for view synthesis , author=. Communications of the ACM , volume=. 2021 , publisher=

2021
[18]

, author=

3D Gaussian splatting for real-time radiance field rendering. , author=. ACM Trans. Graph. , volume=
[19]

Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

Midi: Multi-instance diffusion for single image to 3d scene generation , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=
[20]

arXiv preprint arXiv:2507.13346 , year=

AutoPartGen: Autogressive 3D Part Generation and Discovery , author=. arXiv preprint arXiv:2507.13346 , year=

arXiv
[21]

arXiv preprint arXiv:2507.08772 , year=

From One to More: Contextual Part Latents for 3D Generation , author=. arXiv preprint arXiv:2507.08772 , year=
[22]

ACM Transactions on Graphics (TOG) , volume=

Cast: Component-aligned 3d scene reconstruction from an rgb image , author=. ACM Transactions on Graphics (TOG) , volume=. 2025 , publisher=

2025
[23]

Advances in neural information processing systems , volume=

Denoising diffusion probabilistic models , author=. Advances in neural information processing systems , volume=
[24]

arXiv preprint arXiv:2010.02502 , year=

Denoising diffusion implicit models , author=. arXiv preprint arXiv:2010.02502 , year=

2010 arXiv
[25]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Objaverse: A universe of annotated 3d objects , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[26]

Advances in Neural Information Processing Systems , volume=

Objaverse-xl: A universe of 10m+ 3d objects , author=. Advances in Neural Information Processing Systems , volume=
[27]

Advances in Neural Information Processing Systems , volume=

One-2-3-45: Any single image to 3d mesh in 45 seconds without per-shape optimization , author=. Advances in Neural Information Processing Systems , volume=
[28]

arXiv preprint arXiv:2507.15399 , year=

Blended Point Cloud Diffusion for Localized Text-guided Shape Editing , author=. arXiv preprint arXiv:2507.15399 , year=

arXiv
[29]

arXiv preprint arXiv:2309.03453 , year=

Syncdreamer: Generating multiview-consistent images from a single-view image , author=. arXiv preprint arXiv:2309.03453 , year=

arXiv
[30]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Wonder3d: Single image to 3d using cross-domain diffusion , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[31]

arXiv preprint arXiv:2311.04400 , year=

Lrm: Large reconstruction model for single image to 3d , author=. arXiv preprint arXiv:2311.04400 , year=

arXiv
[32]

European Conference on Computer Vision , pages=

Lgm: Large multi-view gaussian model for high-resolution 3d content creation , author=. European Conference on Computer Vision , pages=. 2025 , organization=

2025
[33]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Epidiff: Enhancing multi-view synthesis via localized epipolar-constrained diffusion , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[34]

ACM Transactions on Graphics (TOG) , volume=

CLAY: A Controllable Large-scale Generative Model for Creating High-quality 3D Assets , author=. ACM Transactions on Graphics (TOG) , volume=. 2024 , publisher=

2024
[35]

arXiv preprint arXiv:2405.20343 , year=

Unique3D: High-Quality and Efficient 3D Mesh Generation from a Single Image , author=. arXiv preprint arXiv:2405.20343 , year=

arXiv
[37]

arXiv preprint arXiv:2406.03184 , year=

Ouroboros3D: Image-to-3D Generation via 3D-aware Recursive Diffusion , author=. arXiv preprint arXiv:2406.03184 , year=

arXiv
[38]

arXiv preprint arXiv:2404.07191 , year=

Instantmesh: Efficient 3d mesh generation from a single image with sparse-view large reconstruction models , author=. arXiv preprint arXiv:2404.07191 , year=

arXiv
[39]

European Conference on Computer Vision , pages=

Sv3d: Novel multi-view synthesis and 3d generation from a single image using latent video diffusion , author=. European Conference on Computer Vision , pages=. 2025 , organization=

2025
[40]

arXiv preprint arXiv:2403.05034 , year=

Crm: Single image to 3d textured mesh with convolutional reconstruction model , author=. arXiv preprint arXiv:2403.05034 , year=

arXiv
[41]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

One-2-3-45++: Fast single image to 3d objects with consistent multi-view generation and 3d diffusion , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[42]

arXiv preprint arXiv:2405.14832 , year=

Direct3D: Scalable Image-to-3D Generation via 3D Latent Diffusion Transformer , author=. arXiv preprint arXiv:2405.14832 , year=

arXiv
[43]

Advances in Neural Information Processing Systems , volume=

Michelangelo: Conditional 3d shape generation based on shape-image-text aligned latent representation , author=. Advances in Neural Information Processing Systems , volume=
[44]

arXiv preprint arXiv:2410.13530 , year=

L3DG: Latent 3D Gaussian Diffusion , author=. arXiv preprint arXiv:2410.13530 , year=

arXiv
[45]

ACM Transactions on Graphics (TOG) , volume=

Blockfusion: Expandable 3d scene generation using latent tri-plane extrapolation , author=. ACM Transactions on Graphics (TOG) , volume=. 2024 , publisher=

2024
[46]

arXiv preprint arXiv:2409.08215 , year=

LT3SD: Latent Trees for 3D Scene Diffusion , author=. arXiv preprint arXiv:2409.08215 , year=

arXiv
[47]

ACM SIGGRAPH 2024 Conference Papers , pages=

Part123: part-aware 3d reconstruction from a single-view image , author=. ACM SIGGRAPH 2024 Conference Papers , pages=

2024
[48]

European Conference on Computer Vision , pages=

Tela: Text to layer-wise 3d clothed human generation , author=. European Conference on Computer Vision , pages=. 2025 , organization=

2025
[49]

2025 , eprint=

CraftsMan3D: High-fidelity Mesh Generation with 3D Native Generation and Interactive Geometry Refiner , author=. 2025 , eprint=

2025
[50]

2024 , eprint=

MeshXL: Neural Coordinate Field for Generative 3D Foundation Models , author=. 2024 , eprint=

2024
[51]

2024 , eprint=

MeshAnything: Artist-Created Mesh Generation with Autoregressive Transformers , author=. 2024 , eprint=

2024
[52]

2024 , eprint=

LLaMA-Mesh: Unifying 3D Mesh Generation with Language Models , author=. 2024 , eprint=

2024
[53]

2024 , eprint=

Meshtron: High-Fidelity, Artist-Like 3D Mesh Generation at Scale , author=. 2024 , eprint=

2024
[54]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Sparseflex: High-resolution and arbitrary-topology 3d shape modeling , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[55]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Dora: Sampling and benchmarking for 3d shape variational auto-encoders , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[56]

2025 , eprint=

Structured 3D Latents for Scalable and Versatile 3D Generation , author=. 2025 , eprint=

2025
[57]

2024 , eprint=

Neural LightRig: Unlocking Accurate Object Normal and Material Estimation with Multi-Light Diffusion , author=. 2024 , eprint=

2024
[58]

2025 , eprint=

MeshArt: Generating Articulated Meshes with Structure-Guided Transformers , author=. 2025 , eprint=

2025
[59]

2025 , eprint=

TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models , author=. 2025 , eprint=

2025
[60]

2025 , eprint=

DeepMesh: Auto-Regressive Artist-mesh Creation with Reinforcement Learning , author=. 2025 , eprint=

2025
[61]

2025 , eprint=

OctGPT: Octree-based Multiscale Autoregressive Models for 3D Shape Generation , author=. 2025 , eprint=

2025
[62]

2025 , eprint=

Step1X-3D: Towards High-Fidelity and Controllable Generation of Textured 3D Assets , author=. 2025 , eprint=

2025
[63]

2025 , eprint=

Direct3D-S2: Gigascale 3D Generation Made Easy with Spatial Sparse Attention , author=. 2025 , eprint=

2025
[64]

2025 , eprint=

DIPO: Dual-State Images Controlled Articulated Object Generation Powered by Diverse Data , author=. 2025 , eprint=

2025
[65]

2025 , eprint=

ShapeLLM-Omni: A Native Multimodal LLM for 3D Generation and Understanding , author=. 2025 , eprint=

2025
[66]

2025 , eprint=

PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers , author=. 2025 , eprint=

2025
[67]

2025 , eprint=

Efficient Part-level 3D Object Generation via Dual Volume Packing , author=. 2025 , eprint=

2025
[68]

2025 , eprint=

Assembler: Scalable 3D Part Assembly via Anchor Point Diffusion , author=. 2025 , eprint=

2025
[69]

2025 , eprint=

DeOcc-1-to-3: 3D De-Occlusion from a Single Image via Self-Supervised Multi-View Diffusion , author=. 2025 , eprint=

2025
[70]

2025 , eprint=

From One to More: Contextual Part Latents for 3D Generation , author=. 2025 , eprint=

2025
[71]

2025 , eprint=

Stereo-GS: Multi-View Stereo Vision Model for Generalizable 3D Gaussian Splatting Reconstruction , author=. 2025 , eprint=

2025
[72]

2025 , eprint=

Ultra3D: Efficient and High-Fidelity 3D Generation with Part Attention , author=. 2025 , eprint=

2025
[73]

arXiv preprint arXiv:1312.6114 , year=

Auto-encoding variational bayes , author=. arXiv preprint arXiv:1312.6114 , year=

arXiv
[74]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Scalable diffusion models with transformers , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[75]

arXiv preprint arXiv:2502.06608 , year=

TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models , author=. arXiv preprint arXiv:2502.06608 , year=

arXiv
[76]

arXiv preprint arXiv:2412.03632 , year=

Mv-adapter: Multi-view consistent image generation made easy , author=. arXiv preprint arXiv:2412.03632 , year=

arXiv
[77]

2022 , eprint=

DreamFusion: Text-to-3D using 2D Diffusion , author=. 2022 , eprint=

2022
[78]

2023 , eprint=

SHAP-EDITOR: Instruction-guided Latent 3D Editing in Seconds , author=. 2023 , eprint=

2023
[79]

2023 , eprint=

Vox-E: Text-guided Voxel Editing of 3D Objects , author=. 2023 , eprint=

2023
[80]

2024 , eprint=

TIP-Editor: An Accurate 3D Editor Following Both Text-Prompts And Image-Prompts , author=. 2024 , eprint=

2024
[81]

2024 , eprint=

Generic 3D Diffusion Adapter Using Controlled Multi-View Editing , author=. 2024 , eprint=

2024
[82]

2024 , eprint=

Make-Your-3D: Fast and Consistent Subject-Driven 3D Content Generation , author=. 2024 , eprint=

2024
[83]

2024 , eprint=

Interactive3D: Create What You Want by Interactive 3D Generation , author=. 2024 , eprint=

2024
[84]

2024 , eprint=

Tailor3D: Customized 3D Assets Editing and Generation with Dual-Side Images , author=. 2024 , eprint=

2024
[85]

2024 , eprint=

MVInpainter: Learning Multi-View Consistent Inpainting to Bridge 2D and 3D Editing , author=. 2024 , eprint=

2024
[86]

2025 , eprint=

The Scene Language: Representing Scenes with Programs, Words, and Embeddings , author=. 2025 , eprint=

2025
[87]

2024 , eprint=

PrEditor3D: Fast and Precise 3D Shape Editing , author=. 2024 , eprint=

2024
[88]

2024 , eprint=

3D Mesh Editing using Masked LRMs , author=. 2024 , eprint=

2024
[89]

arXiv preprint arXiv:2308.16512 , year=

Mvdream: Multi-view diffusion for 3d generation , author=. arXiv preprint arXiv:2308.16512 , year=

arXiv
[90]

2025 , eprint=

MeshPad: Interactive Sketch-Conditioned Artist-Reminiscent Mesh Generation and Editing , author=. 2025 , eprint=

2025
[91]

2025 , eprint=

Geometry in Style: 3D Stylization via Surface Normal Deformation , author=. 2025 , eprint=

2025
[92]

2025 , eprint=

CMD: Controllable Multiview Diffusion for 3D Editing and Progressive Generation , author=. 2025 , eprint=

2025
[93]

ACM Transactions On Graphics (TOG) , volume=

3dshape2vecset: A 3d shape representation for neural fields and generative diffusion models , author=. ACM Transactions On Graphics (TOG) , volume=. 2023 , publisher=

2023
[94]

arXiv preprint arXiv:2505.14521 , year=

Sparc3D: Sparse Representation and Construction for High-Resolution 3D Shapes Modeling , author=. arXiv preprint arXiv:2505.14521 , year=

arXiv
[95]

2025 , eprint=

Pro3D-Editor : A Progressive-Views Perspective for Consistent and Precise 3D Editing , author=. 2025 , eprint=

2025
[96]

2025 , eprint=

EditP23: 3D Editing via Propagation of Image Prompts to Multi-View , author=. 2025 , eprint=

2025
[97]

2025 , eprint=

Blended Point Cloud Diffusion for Localized Text-guided Shape Editing , author=. 2025 , eprint=

2025
[98]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

An edit friendly ddpm noise space: Inversion and manipulations , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[99]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Null-text inversion for editing real images using guided diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[100]

arXiv preprint arXiv:2204.06125 , volume=

Hierarchical text-conditional image generation with clip latents , author=. arXiv preprint arXiv:2204.06125 , volume=

arXiv
[101]

Advances in neural information processing systems , volume=

Photorealistic text-to-image diffusion models with deep language understanding , author=. Advances in neural information processing systems , volume=
[102]

2022 , eprint=

High-Resolution Image Synthesis with Latent Diffusion Models , author=. 2022 , eprint=

2022
[103]

2024 , eprint=

Scaling Rectified Flow Transformers for High-Resolution Image Synthesis , author=. 2024 , eprint=

2024
[104]

arXiv preprint arXiv:2410.10792 , year=

Semantic image inversion and editing using rectified stochastic differential equations , author=. arXiv preprint arXiv:2410.10792 , year=

arXiv
[105]

arXiv preprint arXiv:2411.04746 , year=

Taming rectified flow for inversion and editing , author=. arXiv preprint arXiv:2411.04746 , year=

arXiv
[106]

2022 , eprint=

An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion , author=. 2022 , eprint=

2022
[107]

2025 , eprint=

KV-Edit: Training-Free Image Editing for Precise Background Preservation , author=. 2025 , eprint=

2025
[108]

2023 , eprint=

Prompt Tuning Inversion for Text-Driven Image Editing Using Diffusion Models , author=. 2023 , eprint=

2023
[109]

2025 , eprint=

Personalize Anything for Free with Diffusion Transformer , author=. 2025 , eprint=

2025
[110]

2025 , eprint=

Amodal3R: Amodal 3D Reconstruction from Occluded 2D Images , author=. 2025 , eprint=

2025
[111]

2025 , eprint=

FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space , author=. 2025 , eprint=

2025
[112]

2024 , howpublished=

Black Forest Labs , title=. 2024 , howpublished=

2024
[113]

2022 , eprint=

RePaint: Inpainting using Denoising Diffusion Probabilistic Models , author=. 2022 , eprint=

2022
[114]

2022 , eprint=

Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items , author=. 2022 , eprint=

2022
[115]

arXiv preprint arXiv:2509.06784 , year=

P3-sam: Native 3d part segmentation , author=. arXiv preprint arXiv:2509.06784 , year=
[116]

2024 , eprint=

SAMPart3D: Segment Any Part in 3D Objects , author=. 2024 , eprint=

2024
[117]

2024 , month = nov, day =

2024
[118]

2016 , eprint=

A Point Set Generation Network for 3D Object Reconstruction from a Single Image , author=. 2016 , eprint=

2016
[119]

2016 , eprint=

3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction , author=. 2016 , eprint=

2016
[120]

Electronics letters , volume=

Scope of validity of PSNR in image/video quality assessment , author=. Electronics letters , volume=. 2008 , publisher=

2008
[121]

and Sheikh, H.R

Zhou Wang and Bovik, A.C. and Sheikh, H.R. and Simoncelli, E.P. , journal=. Image quality assessment: from error visibility to structural similarity , year=
[122]

2018 , eprint=

The Unreasonable Effectiveness of Deep Features as a Perceptual Metric , author=. 2018 , eprint=

2018
[123]

2024 , eprint=

DINOv2: Learning Robust Visual Features without Supervision , author=. 2024 , eprint=

2024
[124]

2021 , eprint=

Learning Transferable Visual Models From Natural Language Supervision , author=. 2021 , eprint=

2021
[125]

2018 , eprint=

GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium , author=. 2018 , eprint=

2018
[126]

BANG: Dividing 3D Assets via Generative Exploded Dynamics , volume=

Zhang, Longwen and Zhang, Qixuan and Jiang, Haoran and Bai, Yinuo and Yang, Wei and Xu, Lan and Yu, Jingyi , year=. BANG: Dividing 3D Assets via Generative Exploded Dynamics , volume=. ACM Transactions on Graphics , publisher=. doi:10.1145/3730840 , number=

doi
[127]

arXiv preprint arXiv:2508.19247 , year=

Voxhammer: Training-free precise and coherent 3d editing in native 3d space , author=. arXiv preprint arXiv:2508.19247 , year=

arXiv
[128]

arXiv preprint arXiv:2209.14988 , year=

Dreamfusion: Text-to-3d using 2d diffusion , author=. arXiv preprint arXiv:2209.14988 , year=

arXiv
[129]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Vox-e: Text-guided voxel editing of 3d objects , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[130]

Proceedings of the AAAI conference on artificial intelligence , volume=

Focaldreamer: Text-driven 3d editing via focal-fusion assembly , author=. Proceedings of the AAAI conference on artificial intelligence , volume=
[131]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Flowchef: Steering of rectified flow models for controlled generations , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[132]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Flowedit: Inversion-free text-based editing using pre-trained flow models , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[133]

Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

Alias-free latent diffusion models: Improving fractional shift equivariance of diffusion latent space , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=
[134]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Zero-1-to-3: Zero-shot one image to 3d object , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[135]

arXiv preprint arXiv:2510.20776 , year=

CUPID: Generative 3D Reconstruction via Joint Object and Pose Modeling , author=. arXiv preprint arXiv:2510.20776 , year=
[136]

Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

Instant3dit: Multiview inpainting for fast editing of 3d objects , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=
[137]

arXiv preprint arXiv:2209.03003 , year=

Flow straight and fast: Learning to generate and transfer data with rectified flow , author=. arXiv preprint arXiv:2209.03003 , year=

arXiv
[138]

Gemini 3 Pro: A Multimodal Large Language Model , year =
[139]

arXiv preprint arXiv:2510.15019 , year=

NANO3D: A Training-Free Approach for Efficient 3D Editing Without Masks , author=. arXiv preprint arXiv:2510.15019 , year=
[140]

arXiv preprint arXiv:2208.01626 , year=

Prompt-to-prompt image editing with cross attention control , author=. arXiv preprint arXiv:2208.01626 , year=

arXiv
[141]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Instructpix2pix: Learning to follow image editing instructions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[142]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Magic3d: High-resolution text-to-3d content creation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[143]

Advances in neural information processing systems , volume=

Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation , author=. Advances in neural information processing systems , volume=
[144]

arXiv preprint arXiv:2212.08751 , year=

Point-e: A system for generating 3d point clouds from complex prompts , author=. arXiv preprint arXiv:2212.08751 , year=

arXiv
[145]

arXiv preprint arXiv:2305.02463 , year=

Shap-e: Generating conditional 3d implicit functions , author=. arXiv preprint arXiv:2305.02463 , year=

arXiv
[146]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Instruct-nerf2nerf: Editing 3d scenes with instructions , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[147]

SIGGRAPH Asia 2023 conference papers , pages=

Dreameditor: Text-driven 3d scene editing with neural fields , author=. SIGGRAPH Asia 2023 conference papers , pages=

2023
[148]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Score jacobian chaining: Lifting pretrained 2d diffusion models for 3d generation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[149]

Proceedings of the IEEE/CVF international conference on computer vision , pages=

Fantasia3d: Disentangling geometry and appearance for high-quality text-to-3d content creation , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=
[150]

arXiv preprint arXiv:2403.02151 , year=

Triposr: Fast 3d object reconstruction from a single image , author=. arXiv preprint arXiv:2403.02151 , year=

arXiv
[151]

European conference on computer vision , pages=

Crm: Single image to 3d textured mesh with convolutional reconstruction model , author=. European conference on computer vision , pages=. 2024 , organization=

2024
[152]

arXiv preprint arXiv:2405.14979 , year=

Craftsman3d: High-fidelity mesh generation with 3d native generation and interactive geometry refiner , author=. arXiv preprint arXiv:2405.14979 , year=

arXiv
[153]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Gaussianeditor: Swift and controllable 3d editing with gaussian splatting , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[154]

arXiv preprint arXiv:2303.15780 , year=

Instruct 3d-to-3d: Text instruction guided 3d-to-3d conversion , author=. arXiv preprint arXiv:2303.15780 , year=

arXiv
[155]

arXiv preprint arXiv:2605.27351 , year=

Feedforward 3D Editing Learns from Semantic-Part Transformation , author=. arXiv preprint arXiv:2605.27351 , year=

arXiv
[156]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

Easy3E: Feed-Forward 3D Asset Editing via Rectified Voxel Flow , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=
[157]

arXiv preprint arXiv:2501.12202 , year=

Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation , author=. arXiv preprint arXiv:2501.12202 , year=

arXiv
[158]

International conference on learning representations , volume=

Mvdream: Multi-view diffusion for 3d generation , author=. International conference on learning representations , volume=
[160]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=

PrEditor3D: Fast and Precise 3D Shape Editing , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , year=
[161]

arXiv preprint arXiv:2506.15442 , year=

Hunyuan3D 2.1: From Images to High-Fidelity 3D Assets with Production-Ready PBR Material , author=. arXiv preprint arXiv:2506.15442 , year=

arXiv
[162]

arXiv preprint arXiv:2506.00512 , year=

Pro3D-Editor: A Progressive-Views Perspective for Consistent and Precise 3D Editing , author=. arXiv preprint arXiv:2506.00512 , year=

arXiv
[163]

arXiv preprint arXiv:2505.07747 , year=

Step1X-3D: Towards High-Fidelity and Controllable Generation of Textured 3D Assets , author=. arXiv preprint arXiv:2505.07747 , year=

arXiv
[164]

arXiv preprint arXiv:2506.20652 , year=

Editp23: 3d editing via propagation of image prompts to multi-view , author=. arXiv preprint arXiv:2506.20652 , year=

arXiv
[165]

arXiv preprint arXiv:2602.19542 , year=

Vinedresser3D: Agentic Text-guided 3D Editing , author=. arXiv preprint arXiv:2602.19542 , year=
[166]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Gaussianeditor: Editing 3d gaussians delicately with text instructions , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[167]

arXiv preprint arXiv:2108.01073 , year=

Sdedit: Guided image synthesis and editing with stochastic differential equations , author=. arXiv preprint arXiv:2108.01073 , year=

arXiv
[168]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Null-text inversion for editing real images using guided diffusion models , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[169]

1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space , author=

FLUX. 1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space , author=. arXiv preprint arXiv:2506.15742 , year=

arXiv
[170]

arXiv preprint arXiv:2411.07184 , year=

Sampart3d: Segment any part in 3d objects , author=. arXiv preprint arXiv:2411.07184 , year=

arXiv
[171]

Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

Partfield: Learning 3d feature fields for part segmentation and beyond , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=
[172]

arXiv preprint arXiv:2511.13647 , year=

Part-x-mllm: Part-aware 3d multimodal large language model , author=. arXiv preprint arXiv:2511.13647 , year=
[173]

Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

Structured 3d latents for scalable and versatile 3d generation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=
[174]

arXiv preprint arXiv:2204.11918 , year=

Google scanned objects: A high-quality dataset of 3d scanned household items , author=. arXiv preprint arXiv:2204.11918 , year=

arXiv
[175]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=

ABO: Dataset and Benchmarks for Real-World 3D Object Understanding , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , year=
[176]

Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =

Su, Han and Huang, Tianyu and Wan, Zichen and Wu, Xiaohe and Zuo, Wangmeng , title =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , month =. 2026 , pages =

2026
[177]

arXiv preprint arXiv:2511.21631 , year=

Qwen3-VL Technical Report , author=. arXiv preprint arXiv:2511.21631 , year=

arXiv

Pith tools

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

[1] [1]

Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education

Clancey, William J. Communication, Simulation, and Intelligent Agents: Implications of Personal Intelligent Machines for Medical Education. Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI-83)

[2] [2]

Classification Problem Solving

Clancey, William J. Classification Problem Solving. Proceedings of the Fourth National Conference on Artificial Intelligence

[3] [3]

, title =

Robinson, Arthur L. , title =. 1980 , doi =. https://science.sciencemag.org/content/208/4447/1019.full.pdf , journal =

1980

[4] [4]

New Ways to Make Microcircuits Smaller---Duplicate Entry

Robinson, Arthur L. New Ways to Make Microcircuits Smaller---Duplicate Entry. Science

[5] [5]

Clancey and Glenn Rennels , abstract =

Diane Warner Hasling and William J. Clancey and Glenn Rennels , abstract =. Strategic explanations for a diagnostic consultation system , journal =. 1984 , issn =. doi:https://doi.org/10.1016/S0020-7373(84)80003-6 , url =

doi:10.1016/s0020-7373(84)80003-6 1984

[6] [6]

and Rennels, Glenn R

Hasling, Diane Warner and Clancey, William J. and Rennels, Glenn R. and Test, Thomas. Strategic Explanations in Consultation---Duplicate. The International Journal of Man-Machine Studies

[7] [7]

Poligon: A System for Parallel Problem Solving

Rice, James. Poligon: A System for Parallel Problem Solving

[8] [8]

Transfer of Rule-Based Expertise through a Tutorial Dialogue

Clancey, William J. Transfer of Rule-Based Expertise through a Tutorial Dialogue