Pith. sign in

REVIEW 3 major objections 5 minor 59 references

TRACE-GS: On-Policy Trajectory Distillation with Privileged Geometric Conditioning for Sparse-View 3DGS Restoration

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Supervising sparse-view 3DGS diffusion at the states it actually visits, with targets from a dense-view teacher, reduces error accumulation and lifts restoration quality.

desk verdict A solid, well-scoped training-time recipe for sparse-view 3DGS restoration; the on-policy/privileged-geometry combination is new and the ablations are well-designed, but the teacher-reliability question needs direct measurement before the mechanism is fully established. read the letter →

arxiv 2608.10286 v1 pith:7S6LMPKX submitted 2026-08-10 cs.CV

Linlian Jiang , Yuchen Xi , Sadman Rakib Pinon , Ruigang Yang , Yang Wang , Xinxin Zuo This is my paper · ORCID
classification cs.CV
keywords sparse-view3DGaussiansplattingon-policytrajectorydistillationprivilegedinformationdiffusion-basedrestorationoff-policymismatchLoRAadaptationnovelviewsynthesis3DGSrefinement
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

TRACE-GS argues that diffusion-based sparse-view 3DGS restoration fails less from architecture than from where supervision is applied: training at independently noised states leaves the student's own denoising states unsupervised, and under-constrained sparse geometry makes errors compound along the rollout. To fix this, it trains an identical-capacity teacher conditioned on dense-view renderings (geometry available only during training) and queries that frozen teacher at each state the sparse-view student visits, aligning denoising velocities and cross-view retrieval responses there. Only the sparse-view student is deployed, so the richer geometry costs nothing at inference. On DL3DV-Benchmark, Mip-NeRF 360, and NeRFBusters, the paper reports consistent PSNR, SSIM, and LPIPS gains under 3, 6, and 9 input views, with the largest gains at 3 views. The paper's central point is that coupling reliable targets with student-visited states is the transferable ingredient, not a more sophisticated restoration network.

What carries the argument

The load-bearing object is a geometry-asymmetric teacher–student pair built on a frozen video-diffusion restoration backbone with reference cross-attention. The two roles share the frozen weights and rank-32 LoRA adapters cloned from a shared Stage-1 alignment; the teacher is conditioned on dense-view renderings $R^d$ and the student on sparse-view renderings $R^s$, so the asymmetry lies entirely in the conditioning geometry. During on-policy trajectory distillation, the student rolls out a resampled 10-step reverse schedule from noise, the scheduler transitions are detached, and the frozen teacher is queried at each visited state $z^S_{t'_i}$ to align velocity predictions (loss $\mathcal{L}_{\mathrm{traj}}$) and retrieval responses from the final-block reference cross-attention (loss $\mathcal{L}_{\mathrm{ret}}$). Detaching the transitions lets gradients reach the student adapter at every visited state without backpropagating through the rollout, and the teacher's targets are applied exactly where the student needs them. Deployment keeps only the student adapter and feeds its restored, posed pseudo-observations back into iterative 3DGS optimization.

What would settle it

A direct check is to record the teacher's velocity error against ground-truth targets at the student-visited states used by TRACE-GS and compare it with the teacher's error at independently forward-noised states; if the on-policy states are not supervised more accurately than the off-policy states, the claimed benefit of trajectory distillation would be expected to disappear or reverse.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that supervision at independently forward-noised states does not cover the states reached during inference, and in sparse-view 3DGS this off-policy mismatch interacts with geometry-induced bias to accumulate errors along the denoising rollout. TRACE-GS therefore performs on-policy trajectory distillation: a dense-view-conditioned teacher, with the same frozen diffusion backbone and the same capacity as the student and differing only in conditioning, is queried at the sparse-view student's own rollout states to supply velocity targets and normalized reference cross-attention responses, through losses $\mathcal{L}_{\mathrm{traj}}$ and $\mathcal{L}_{\mathrm{ret}}$. The teacher's advantage comes only from renderings of a 3DGS fitted with more training views, putting the method in the learning-using-privileged-information setting; at deployment the student alone restores renderings that serve as pseudo-observations for 3DGS refinement. The paper reports that this consistently improves PSNR, SSIM, and LPIPS on DL3DV-Benchmark, Mip-NeRF 360, and NeRFBusters, with the largest gains under 3-view input and a late-rollout gap between the on-policy and off-policy variants that the authors interpret as error accumulating at uncovered states.

Load-bearing premise

The load-bearing premise is that the dense-view teacher supplies reliable targets at the states the sparse-view student actually visits, even though those states come from a weaker model's rollout and the paper applies no confidence or uncertainty filtering to the teacher's pseudo-observations.

Editorial extensions

If this is right

  • If the central claim holds, training-time dense views can transfer geometric reliability to a deployment-time sparse-view model at zero extra inference cost: teacher, dense reconstruction, and additional views are all discarded at test time.
  • The largest gains under 3 input views follow directly: the sparser the conditioning geometry, the wider the teacher–student gap and the more value on-policy supervision has.
  • Because the on-policy and off-policy variants separate only in the later denoising steps, the method implies that error accumulation, not capacity, is the bottleneck in sparse-view restoration, so later denoising states deserve the most supervision.
  • Retrieval alignment contributes beyond velocity matching, suggesting that constraining how the student attends to reference views along its own trajectory is a usable lever for cross-view consistency.
  • The framework is a drop-in refinement stage for any restore–refine pipeline where extra training views exist, not just the GSFixer-style backbone used here.

Reading between the lines

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

  • Editorial inference: the same on-policy-distillation mechanism should transfer to other conditioning asymmetries, such as depth or multi-modal input, whenever a privileged signal exists only during training; the essential step is querying the privileged model at the deployable model's visited states.
  • Editorial inference: since the teacher is queried without any confidence or uncertainty filtering, a natural strengthening is to weight on-policy targets by estimated teacher error, and a failure analysis should track whether teacher error at visited states is the limiting factor.
  • Editorial inference: the per-step results imply that later rollout states carry most of the signal, so annealing the Stage-2 rollout length (short early, long late) is a testable way to cut memory cost without losing the late-state benefit.
  • Editorial inference: the off-policy variant's per-step plateau suggests a general lesson for diffusion restoration beyond 3DGS: replaying real student states during training is valuable whenever the restoration input itself changes across refinement rounds.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces TRACE-GS, a training-time distillation framework for sparse-view 3D Gaussian Splatting restoration. The method freezes a video-diffusion restoration backbone and trains LoRA adapters so that a dense-view-conditioned teacher and a sparse-view-conditioned student share the same architecture and capacity. After a direction-alignment warm-up, the student performs its own reverse-diffusion rollout, and the frozen teacher is queried at each student-visited state, providing trajectory-distillation and retrieval-alignment targets. At deployment only the sparse-conditioned student adapter is retained and its restored renderings are used as pseudo-observations in an iterative 3DGS refinement loop. Experiments on DL3DV-Benchmark, Mip-NeRF 360, and NeRFBusters report consistent PSNR/SSIM/LPIPS gains over diffusion-based baselines under 3, 6, and 9 input views, and ablations isolate the effect of privileged geometry and of on-policy versus off-policy state selection.

Significance. If the central claim is validated, the paper makes a useful conceptual contribution: it identifies the training-inference state mismatch in diffusion-based sparse-view 3DGS restoration and addresses it with a LUPI-style dense-conditioned teacher, without adding deployment-time cost. The strengths of the paper are the clear problem formulation, the capacity-matched teacher-student design, and the well-constructed ablation (E) versus (F) that isolates on-policy state selection from training budget. The deployment procedure is carefully matched to GSFixer, and the authors state explicit limitations. The main weakness is empirical verification: all quantitative claims rest on single-run means, several baseline numbers are copied from a previous evaluation, and the reliability of teacher targets at student-visited states is never measured. These issues are fixable and do not invalidate the idea, but they currently prevent the central claim from being considered established.

major comments (3)
  1. [Sec. 3.4 / Eqs. (5)-(8)] The load-bearing premise that the dense-conditioned teacher supplies reliable targets at student-visited states is asserted but never measured. The teacher is trained on forward-noised states derived from offline pseudo-targets (Sec. A.4), whereas at Stage 2 it is queried at states produced by the sparse-conditioned student under a detached scheduler; these states can lie outside the teacher's training distribution. Sec. A.5 explicitly applies no uncertainty or confidence filtering, and the Limitations section relegates uncertainty-aware supervision to future work. Because the Table 3 (E)-versus-(F) comparison shows only net gains (+0.17 dB on DL3DV, +0.31 dB on Mip-NeRF 360), the observed improvement could in principle come from additional state coverage or from implicit regularization rather than from the reliability of privileged targets. Please provide a diagnostic of teacher error at visited states, for example the teacher velocity error against the warm-up reference at matched timesteps or against ground-truth clean latents where available, and report how this error evolves along the rollout. This is needed to substantiate the mechanism claimed in Eqs. (5)-(8).
  2. [Sec. 4.2 / Tables 1-3] All reported numbers are single-run means without error bars, standard deviations, or significance tests. Several decisive margins are small: in Table 3, variant (E) versus (F) differs by 0.17 dB PSNR on DL3DV and 0.31 dB on Mip-NeRF 360; in Table 2, the 9-view PSNR gain over GSFixer is at most 0.57 dB. Without a measure of run-to-run variance, the ordering of variants and the claim of consistent gains are not statistically established. Please report results over at least three seeds for the main comparisons and the core ablations, or provide confidence intervals, and state whether the same pseudo-targets and evaluation scenes are reused across seeds.
  3. [Sec. B / Tables 1-2] A substantial fraction of the baseline numbers in Tables 1 and 2 is taken from the GSFixer common evaluation rather than re-run under the exact protocol used for TRACE-GS. The supplement states that baseline values from GSFixer are reported unchanged. Because the central empirical claim rests on comparison with these baselines, and because the restore-refine pipeline (trajectory generation, repair schedule, pseudo-observation weighting) can change results by amounts comparable to the reported gaps, the paper should clearly mark which cells are copied, verify protocol equivalence, and ideally re-run at least the closest baseline (GSFixer) in the same code environment. Without this, the headline improvements over prior methods are not fully verifiable.
minor comments (5)
  1. [Eq. (2)] Equation (2) contains underbrace annotations that are difficult to read in the submitted PDF; please typeset the teacher-role and student-role annotations more cleanly.
  2. [Sec. 4.1] The relationship between training-time sparse renderings taken at 7K and 17K iterations and the deployment repair schedule at 7K, 17K, and 27K iterations is not explained; clarify whether the two sets of iteration numbers are intended to correspond.
  3. [Figure 6] The caption does not state how many clips are used for the per-step evaluation or whether intermediate PSNR/LPIPS are computed on decoded latents or on features; please provide these details in the caption or in the supplement.
  4. [Eq. (7)] The definition of H_i^b is given after the equation that uses it; reorder the presentation so that H_i^b is defined before the attention expression.
  5. [Tables 1-2] The phrase 'colors from yellow to red' is not accessible in grayscale printing; consider adding numeric emphasis or a separate table note.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: teacher targets are externally conditioned on dense-view geometry, and the student-visited-state supervision is a genuine on-policy distillation objective.

full rationale

The derivation chain is self-contained and no load-bearing step reduces to its own input. The central objective L_traj = E[||uS_i - sg[uT_i]||^2] (Eq. 6) compares the student's velocity at its own detached rollout states against the frozen teacher's velocity at those same states; the teacher is conditioned on dense-view renderings Rd and is not fitted to the student's outputs. Likewise, L_ret (Eq. 8) aligns retrieval responses against the same frozen teacher, and the dense renderings and pseudo-targets bIT are produced offline by a separate pretrained restorer from dense-view reconstructions (Sec. A.4), not by the student. No fitted parameter is renamed as a prediction: Tables 1-2 report held-out and out-of-domain benchmarks under a matched evaluation protocol. The cited off-policy mismatch and on-policy imitation concepts come from external prior work (Ross et al. 2011; Li et al. 2024b; Ning et al. 2024), not from a self-citation chain, and no uniqueness theorem or ansatz is imported from the authors' own prior work. The paper's admitted lack of teacher-error or confidence filtering at student-visited states (Sec. A.5 and Limitations) is a correctness/validity risk rather than a circularity; it concerns whether the teacher's targets are accurate, not whether the objective is definitionally equivalent to its inputs.

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

The method introduces no new physical or architectural entities; it combines existing components (frozen video diffusion, LoRA adapters, 3DGS, 3DGS renderings) with new training objectives. The main load-bearing assumptions are domain assumptions about geometry reliability, teacher reliability, and pseudo-target quality. The central free parameters are hyperparameters fixed by hand without sensitivity analysis.

free parameters (5)
  • Lambda_ret = 0.15
    Weight for on-policy retrieval alignment in Eq. (9). Chosen by hand with no reported sensitivity analysis.
  • Lambda_align = 1.0
    Weight for Stage-1 privileged direction alignment in Eq. (16). No ablation is reported for this value.
  • Rollout steps N = 10
    Number of student-visited states per rollout in Sec. 3.4 and Alg. 1. Deployment uses M=50 steps; no sensitivity analysis for N.
  • Repair schedule and rounds = 3 rounds; J_repair = {7000, 17000, 27000}; lambda_reg = 1
    Deployment-time refinement protocol in Sec. A.5. These values are fixed by hand and not ablated.
  • LoRA rank r = 32
    Adapter rank used for both teacher and student. No ablation of rank is provided.
assumptions (4)
  • domain assumption Dense-view renderings R_d carry more reliable geometry than sparse-view renderings R_s.
    Sec. 3.2 motivates privileged geometry with this premise. If dense renderings are not more reliable, the teacher has no advantage to transfer.
  • domain assumption The frozen teacher provides reliable velocity and retrieval targets at student-visited states.
    Eqs. (5)-(8) use teacher outputs as supervision at on-policy states. The paper does not measure teacher error at these states and applies no confidence filtering.
  • domain assumption Pseudo-targets I_T, generated by a frozen GSFixer from R_d, are a valid training target.
    Sec. A.4 uses offline GSFixer-generated restorations as flow-matching targets. Any artifacts in I_T are inherited by both teacher and student.
  • ad hoc to paper The Stage-1 shared adapter, trained with L_warm, provides a sufficient initialization for Stage-2 distillation.
    The two-stage schedule in Sec. A.4 is presented without evidence that another initialization would produce similar results.

how reviews work

0 comments
Cite this review

Pith. "Pith review of TRACE-GS: On-Policy Trajectory Distillation with Privileged Geometric Conditioning for Sparse-View 3DGS Restoration." pith.science (2026). https://pith.science/paper/7S6LMPKX

@misc{pith2026260810286,
  author       = {Pith},
  title        = {Pith review of: TRACE-GS: On-Policy Trajectory Distillation with Privileged Geometric Conditioning for Sparse-View 3DGS Restoration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7S6LMPKX}},
  note         = {Machine review of arXiv:2608.10286}
}
read the original abstract

We present TRACE-GS, an on-policy trajectory distillation framework that leverages privileged geometric conditioning at training time, thereby adapting a diffusion prior to sparse-view 3D Gaussian Splatting (3DGS) restoration. Rather than pursuing increasingly sophisticated restoration architectures, we identify a more fundamental limitation shared by existing diffusion-based approaches: supervision at independently noised states does not cover those reached during inference. In sparse-view 3DGS, under-constrained geometry biases denoising from the outset, and the resulting deviations compound along the rollout. TRACE-GS instead performs on-policy trajectory distillation: a teacher conditioned on richer geometry from additional training views supplies targets along the sparse-view student's own rollout, aligning denoising directions and cross-view responses at each visited state. This training-only geometry places TRACE-GS in the learning using privileged information (LUPI) setting. At deployment, only the sparse-view student is retained, and its restored renderings serve as pseudo-observations for 3DGS refinement. To the best of our knowledge, TRACE-GS is the first to derive on-policy supervision from privileged geometry for sparse-view 3DGS restoration, achieving consistent gains and strong generalization across datasets and sparse-view settings.

Figures

Figures reproduced from arXiv: 2608.10286 by the authors.

Figure 1
Figure 1. Off-policy mismatch in sparse-view 3DGS restora [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of TRACE-GS. (A) Given Vs ⊂ Vd, we independently reconstruct a 3DGS from each view set and render paired sequences Rs and Rd along the same target trajectory. (B) Built on the same frozen video-diffusion backbone, the teacher and student use a frozen LoRA ϕT and a trainable LoRA ϕS cloned from a shared initialization, and are conditioned on Rd and Rs, respectively, forming a geometry-asymmetric pair. The st… view at source ↗
Figure 3
Figure 3. Qualitative comparison on DL3DV. GSFixer blurs [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative comparison on out-of-domain sparse-view reconstruction using 3, 6, and 9 views on Mip-NeRF 360 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Per-step restoration quality along the reverse dif [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison on a zoomed text region. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Qualitative comparison on out-of-domain sparse-view reconstruction on Mip-NeRF 360 [PITH_FULL_IMAGE:figures/full_fig_p016_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison on out-of-domain sparse-view reconstruction on NeRFBusters [PITH_FULL_IMAGE:figures/full_fig_p017_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

59 extracted references · 19 canonical work pages

  1. [1]

    , author=

    3d gaussian splatting for real-time radiance field rendering. , author=. ACM Trans. Graph. , volume=

  2. [2]

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

    Freenerf: Improving few-shot neural rendering with free frequency regularization , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  3. [3]

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

    Sparsenerf: Distilling depth ranking for few-shot novel view synthesis , author=. Proceedings of the IEEE/CVF international conference on computer vision , pages=

  4. [4]

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

    pixelsplat: 3d gaussian splats from image pairs for scalable generalizable 3d reconstruction , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  5. [5]

    European conference on computer vision , pages=

    Mvsplat: Efficient 3d gaussian splatting from sparse multi-view images , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  6. [6]

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

    Nerfbusters: Removing ghostly artifacts from casually captured nerfs , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

    work page

  7. [7]

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

    Reconfusion: 3d reconstruction with diffusion priors , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  8. [8]

    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=

Show all 59 references
  1. [9]

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

    Difix3d+: Improving 3d reconstructions with single-step diffusion models , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  2. [10]

    arXiv preprint arXiv:2508.09667 , year=

    Gsfixer: Improving 3d gaussian splatting with reference-guided video diffusion priors , author=. arXiv preprint arXiv:2508.09667 , year=

    arXiv

  3. [11]

    arXiv preprint arXiv:2601.18734 , year=

    Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models , author=. arXiv preprint arXiv:2601.18734 , year=

    arXiv

  4. [12]

    arXiv preprint arXiv:2605.05204 , year=

    D-OPSD: On-Policy Self-Distillation for Continuously Tuning Step-Distilled Diffusion Models , author=. arXiv preprint arXiv:2605.05204 , year=

    arXiv

  5. [13]

    arXiv preprint arXiv:2604.08532 , year=

    Self-Improving 4D Perception via Self-Distillation , author=. arXiv preprint arXiv:2604.08532 , year=

    arXiv

  6. [14]

    arXiv preprint arXiv:1503.02531 , year=

    Distilling the knowledge in a neural network , author=. arXiv preprint arXiv:1503.02531 , year=

    arXiv

  7. [15]

    Advances in Neural Information Processing Systems , volume=

    Depth anything v2 , author=. Advances in Neural Information Processing Systems , volume=

  8. [16]

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

    One-step diffusion with distribution matching distillation , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  9. [17]

    Advances in Neural Information Processing Systems , volume=

    Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models , author=. Advances in Neural Information Processing Systems , volume=

  10. [18]

    Forty-first International Conference on Machine Learning , year=

    Score identity distillation: Exponentially fast distillation of pretrained diffusion models for one-step generation , author=. Forty-first International Conference on Machine Learning , year=

  11. [19]

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

    Learning few-step diffusion models by trajectory distribution matching , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  12. [20]

    Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track) , pages=

    Gkd: A general knowledge distillation framework for large-scale pre-trained language model , author=. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track) , pages=

  13. [21]

    Neural networks , volume=

    A new learning paradigm: Learning using privileged information , author=. Neural networks , volume=. 2009 , publisher=

    2009

  14. [22]

    arXiv preprint arXiv:1511.03643 , year=

    Unifying distillation and privileged information , author=. arXiv preprint arXiv:1511.03643 , year=

    arXiv

  15. [23]

    2003 , publisher=

    Information theory, inference and learning algorithms , author=. 2003 , publisher=

    2003

  16. [24]

    Journal of Machine Learning Research , volume=

    gsplat: An open-source library for Gaussian splatting , author=. Journal of Machine Learning Research , volume=

  17. [25]

    arXiv preprint arXiv:2408.06072 , year=

    CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer , author=. arXiv preprint arXiv:2408.06072 , year=

    arXiv

  18. [26]

    ACM Transactions on Graphics (ToG) , volume=

    Tanks and temples: Benchmarking large-scale scene reconstruction , author=. ACM Transactions on Graphics (ToG) , volume=. 2017 , publisher=

    2017

  19. [27]

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

    Mip-nerf 360: Unbounded anti-aliased neural radiance fields , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  20. [28]

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

    Dl3dv-10k: A large-scale scene dataset for deep learning-based 3d vision , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  21. [29]

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

    The unreasonable effectiveness of deep features as a perceptual metric , author=. Proceedings of the IEEE conference on computer vision and pattern recognition , pages=

  22. [30]

    IEEE transactions on image processing , volume=

    Image quality assessment: from error visibility to structural similarity , author=. IEEE transactions on image processing , volume=. 2004 , publisher=

    2004

  23. [31]

    Advances in neural information processing systems , volume=

    Gans trained by a two time-scale update rule converge to a local nash equilibrium , author=. Advances in neural information processing systems , volume=

  24. [32]

    arXiv preprint arXiv:2308.06721 , year=

    Ip-adapter: Text compatible image prompt adapter for text-to-image diffusion models , author=. arXiv preprint arXiv:2308.06721 , year=

    arXiv

  25. [33]

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

    Genfusion: Closing the loop between reconstruction and generation via videos , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  26. [34]

    The Twelfth International Conference on Learning Representations , year=

    A good learner can teach better: Teacher-student collaborative knowledge distillation , author=. The Twelfth International Conference on Learning Representations , year=

  27. [35]

    arXiv preprint arXiv:2503.20314 , year=

    Wan: Open and advanced large-scale video generative models , author=. arXiv preprint arXiv:2503.20314 , year=

    arXiv

  28. [36]

    arXiv preprint arXiv:2301.11706 , year=

    Input perturbation reduces exposure bias in diffusion models , author=. arXiv preprint arXiv:2301.11706 , year=

    arXiv

  29. [37]

    International Conference on Learning Representations , volume=

    Alleviating exposure bias in diffusion models through sampling with shifted time steps , author=. International Conference on Learning Representations , volume=

  30. [38]

    Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=

    A reduction of imitation learning and structured prediction to no-regret online learning , author=. Proceedings of the fourteenth international conference on artificial intelligence and statistics , pages=. 2011 , organization=

    2011

  31. [39]

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

    Splatad: Real-time lidar and camera rendering with 3d gaussian splatting for autonomous driving , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  32. [40]

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

    Scene-Agnostic Object-Centric Representation Learning for 3D Gaussian Splatting , author=. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages=

  33. [41]

    ACM SIGGRAPH 2024 conference papers , pages=

    Vr-gs: A physical dynamics-aware interactive gaussian splatting system in virtual reality , author=. ACM SIGGRAPH 2024 conference papers , pages=

    2024

  34. [42]

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

    Dngaussian: Optimizing sparse-view 3d gaussian radiance fields with global-local depth normalization , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  35. [43]

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

    Nexusgs: Sparse view synthesis with epipolar depth priors in 3d gaussian splatting , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  36. [44]

    European conference on computer vision , pages=

    Cor-gs: sparse-view 3d gaussian splatting via co-regularization , author=. European conference on computer vision , pages=. 2024 , organization=

    2024

  37. [45]

    Proceedings of the computer vision and pattern recognition conference , pages=

    Dropgaussian: Structural regularization for sparse-view gaussian splatting , author=. Proceedings of the computer vision and pattern recognition conference , pages=

  38. [46]

    ACM SIGGRAPH 2024 conference papers , pages=

    2d gaussian splatting for geometrically accurate radiance fields , author=. ACM SIGGRAPH 2024 conference papers , pages=

    2024

  39. [47]

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

    Zip-nerf: Anti-aliased grid-based neural radiance fields , author=. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages=

  40. [48]

    European Conference on Computer Vision , pages=

    Deceptive-nerf/3dgs: Diffusion-generated pseudo-observations for high-quality sparse-view reconstruction , author=. European Conference on Computer Vision , pages=. 2024 , organization=

    2024

  41. [49]

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

    3denhancer: Consistent multi-view diffusion for 3d enhancement , author=. Proceedings of the Computer Vision and Pattern Recognition Conference , pages=

  42. [50]

    International Conference on Learning Representations , volume=

    Elucidating the exposure bias in diffusion models , author=. International Conference on Learning Representations , volume=

  43. [51]

    International Conference on Learning Representations , volume=

    On-policy distillation of language models: Learning from self-generated mistakes , author=. International Conference on Learning Representations , volume=

  44. [52]

    International Conference on Learning Representations , volume=

    Minillm: Knowledge distillation of large language models , author=. International Conference on Learning Representations , volume=

  45. [53]

    arXiv preprint arXiv:2408.14319 , year=

    Rethinking Knowledge Transfer in Learning Using Privileged Information , author=. arXiv preprint arXiv:2408.14319 , year=

    arXiv

  46. [54]

    2026 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=

    OracleGS: Grounding Generative Priors for Sparse-View Gaussian Splatting , author=. 2026 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) , pages=. 2026 , organization=

    2026

  47. [55]

    International Conference on Learning Representations , volume=

    Student-informed teacher training , author=. International Conference on Learning Representations , volume=

  48. [56]

    arXiv preprint arXiv:2402.06379 , year=

    Learning using privileged information for segmenting tumors on digital mammograms , author=. arXiv preprint arXiv:2402.06379 , year=

    arXiv

  49. [57]

    International conference on machine learning , pages=

    Transfer and marginalize: Explaining away label noise with privileged information , author=. International conference on machine learning , pages=. 2022 , organization=

    2022

  50. [58]

    Advances in Neural Information Processing Systems , volume=

    Probabilistic conformal distillation for enhancing missing modality robustness , author=. Advances in Neural Information Processing Systems , volume=

  51. [59]

    arXiv preprint arXiv:2604.12617 , year=

    SOAR: Self-Correction for Optimal Alignment and Refinement in Diffusion Models , author=. arXiv preprint arXiv:2604.12617 , year=

    arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.