Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

Score Change of Variables

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Pith's one-line read A change-of-variables identity for score functions lets diffusion models train in one space and sample in another.

desk verdict GSSM is a genuine extension worth fixing up, but the headline formula is a known identity and the identifiability condition is unstated. read the letter →

arxiv 2412.07904 v3 pith:Z5Y4YYH3 submitted 2024-12-10 cs.LG cs.AImath.PR

classification cs.LGcs.AImath.PR
keywords scorefunctionschangeofvariablesscore-baseddiffusionmodelsreverse-timeItôlemmaslicedmatchinggeneralizedprobabilitysimplexdensityestimation
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 derives a change of variables formula for score functions, the gradients of log-densities that score-based diffusion models estimate. For a smooth invertible map $\phi$, it claims the score in the transformed space equals the pullback of the original score through the inverse Jacobian plus a divergence correction that accounts for the map's curvature. From this identity it proves a reverse-time Itô lemma, so a diffusion process can be trained in one space and reversed in another without relearning the score in the target space, and it builds a generalized sliced score matching objective that replaces random linear projections with gradients of arbitrary smooth functions. A reader should care because this gives a principled way to apply score-based generative models to constrained geometries, such as sampling on a probability simplex, while keeping training in an unconstrained Euclidean space.

What carries the argument

The load-bearing object is the score change-of-variables identity in Theorem 1, an exact formula for how log-density gradients behave under a smooth bijection. It decomposes the transformed score into two pieces: the inverse-Jacobian pullback of the original score, and a divergence term $\nabla_x \cdot \left(J_{\phi^{-1}}(\phi(x))^\top\right)$ evaluated at $x = \phi^{-1}(y)$, which is the score-space analogue of the log-determinant correction in the density change-of-variables formula. The dimension-reduction corollary for $y = v(x)$ with $v: \mathbb{R}^n \to \mathbb{R}$ expresses the scalar score through $\nabla v(x)$, the Laplacian $\Delta v$, and conditional-score terms, and this corollary is what generates the generalized sliced score matching objective. The reverse-time Itô lemma is the same identity transplanted into the standard time-reversal formula for stochastic differential equations.

What would settle it

Train a score network with the un-normalized generalized sliced score matching loss on a two-dimensional mixture of Gaussians, whose true score is known analytically, using the quadratic transformation $v(x) = \tfrac{1}{2} x^\top A x + b^\top x$ with random symmetric $A$. If the recovered vector field differs from the true score at points where $\{Ax + b\}$ does not span $\mathbb{R}^2$, the un-normalized objective is not a faithful score matching loss; if it matches everywhere, the spanning condition is not the binding issue.

Watch

Extended reading notes

Core claim

The central claim is that score functions transform by an exact identity: for $y = \phi(x)$, $\nabla_y \log q(y) = J_{\phi^{-1}}(y)^\top \nabla_x \log p(x) + \nabla_x \cdot \left(J_{\phi^{-1}}(\phi(x))^\top\right)\big|_{x=\phi^{-1}(y)}$. In one dimension this reduces to $(\phi^{-1})'(y)\,\nabla_x \log p(\phi^{-1}(y)) + (\phi^{-1})''(y)/(\phi^{-1})'(y)$. The paper uses this identity to show that reversing the transformed forward SDE yields the same law as transforming the reversed process, so training a score network in $x$-space and sampling in $y$-space is pathwise equivalent to training directly in $y$-space. It also derives a dimension-reduction version for scalar functions $v$, and claims the resulting generalized sliced score matching loss—with extra Hessian and Laplacian terms—is a valid score matching objective that reduces to standard sliced score matching when $v$ is linear.

Load-bearing premise

The paper's generalized sliced score matching objective swaps an exactly weighted loss for an un-normalized one; the two have the same minimizer only if the random projection directions span the whole space at every data point, a condition the paper assumes without proof.

Editorial extensions

If this is right

  • A diffusion model trained in one coordinate system can be sampled in another: the transformed reverse drift is written entirely in terms of the original score, the Jacobian, and the Hessian tensor of $\phi$, with no separate network in the target space.
  • The reverse-time Itô lemma gives a pathwise equivalence: reversing the transformed forward process yields the same law as transforming the reversed process.
  • Generalized sliced score matching strictly generalizes sliced score matching: when $v(x) = v^\top x$ the Hessian and Laplacian terms vanish and the loss reduces to the standard sliced score matching objective.
  • With quadratic random functions, the variance-reduced version has a closed-form quadratic loss for kernel exponential family models, and the reported experiments show lower score-matching loss than variance-reduced sliced score matching on two of three benchmark datasets.
  • On the probability simplex, the additive logistic transformation lets a Gaussian variance-preserving SDE trained in $\mathbb{R}^{12}$ generate valid chess-position representations, with a scalar factor controlling the number of pieces.

Reading between the lines

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

  • The identity would transfer score-based samplers other than diffusion—for example Langevin dynamics or annealed importance sampling—to constrained spaces whenever the map is smooth and invertible and the Jacobian-divergence correction is computable.
  • Because the paper substitutes an un-normalized generalized sliced score matching loss for the exact $1/\|\nabla v\|^4$-weighted score-matching loss, a natural empirical check is to compare the recovered score against an analytically known score in low dimensions; the paper's experiments compare loss values rather than recovered score fields.
  • The scalar factor in the chess experiment acts as a temperature-like control over piece density, so it could be reinterpreted as annealing the target distribution rather than merely rescaling the drift.
  • Learned smooth maps (normalizing flows) could replace hand-derived transformations in the reverse-time lemma, trading analytic Jacobians for automatic differentiation, with Hessian-vector products as the main cost.
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

4 major / 5 minor

Summary. The paper derives a change-of-variables formula for score functions (Theorem 1), uses it to obtain a reverse-time Itô lemma for score-based diffusion models (Corollary 3) and a Generalized Sliced Score Matching method (GSSM). It demonstrates the diffusion result on a chess-position simplex example and GSSM on UCI density estimation tasks. The appendices contain the proofs and additional results.

Significance. If the GSSM identification issue is resolved, the paper offers a useful extension of score matching to nonlinear projections and a theoretically justified method for training in one space and sampling in another. Theorem 1 is a clean identity that could serve as a reference. The paper is honest about limitations and follows established experimental protocols. However, the main new method (GSSM) currently lacks a stated sufficient condition for identifiability, and several derivations in the appendices are not fully rigorous, so the significance is conditional.

major comments (4)
  1. [§C.2, Assumption 2, Eq. (65)] The un-normalized loss (65) is a valid score matching objective only if the matrix M(x) = E_v[∇v(x)∇v(x)^T] is positive definite for p_d-almost every x. Assumption 2 only asserts ∇v(x) ≠ 0 almost surely, which does not imply positive definiteness; for example, if ∇v(x) lies in a fixed proper subspace, M(x) is singular and the loss is constant along the orthogonal complement. The quadratic family in Section 4.2 likely satisfies the stronger condition because b has full-rank covariance, but the paper neither states this condition nor proves it for the quadratic family. Without it, the central GSSM claim that minimizing (65) yields the true score is unsupported.
  2. [§3.2.3, Eq. (14)] Equation (14) contains a missing '+' between the third and fourth terms, and the second term is written as (∇xv)^T ∇_x(sθ ∇xv), which is ambiguous and not what the appendix derives. The correct expression in Eq. (74) is ∇v^T ∇_x sθ ∇v plus a separate Hessian term sθ^T Hv ∇v. As printed, Eq. (14) is not a well-formed objective and must be corrected for reproducibility.
  3. [Appendix A.2, Eq. (32)] The proof of Corollary 2 claims that Eq. (32) is 'trivially true where ∂v/∂x_i = 0', but in that case the right-hand side equals −∂²v/∂x_i², which is generally nonzero. Summing Eq. (32) over all i therefore introduces spurious second-derivative terms for coordinates with zero partial derivative. The proof does not justify the stated formula for v with ∂v/∂x_i = 0 and ∂²v/∂x_i² ≠ 0 at some x. Since Corollary 2 underlies the GSSM derivation, this gap needs to be addressed.
  4. [Appendix A.3, Theorem 2] Theorem 2 is not a valid statement about score functions: if v: R → R^n, then Y = v(X) is supported on a one-dimensional curve in R^n and has no density with respect to Lebesgue measure, so ∇_y log q(y) is not defined in the usual sense. The component-wise treatment ignores the deterministic dependence among the entries of Y. This result should be removed or substantially revised.
minor comments (5)
  1. [§3.2.3] The identification of the linear case would be clearer if the notation explicitly stated that ∇xv is independent of x when v is linear, so that E_v[v^T∇_x(v^T sθ)] reduces to E_v[v^T(∇sθ)v] as in Eq. (13).
  2. [§4.1.4] The claim that increasing w increases the number of pieces is supported only by the qualitative Figure 5; please report a quantitative measure, such as average piece counts, to substantiate this claim.
  3. [Appendix D.4] Test log-likelihoods are reported without error bars or significance tests; given the small UCI datasets, the claim that GSSM-VR outperforms SSM-VR would be stronger with confidence intervals.
  4. [Appendix E.1] There is a typo: 'Guassian' should be 'Gaussian'.
  5. [§5.2] The limitations list acknowledges smoothness and invertibility restrictions, but it could also note that Corollary 2 requires conditional score terms that are not available in practice; the GSSM derivation avoids them only through cancellation, so this should be stated explicitly.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the main theorem is derived from the standard density change-of-variables formula, and the applications are derived rather than fitted.

full rationale

The paper's central claim, Theorem 1, is obtained by taking the gradient of the logarithm of the classical density change-of-variables formula q(y) = p(φ^{-1}(y)) |det J_{φ^{-1}}(y)|. The proof in Appendix A supplies the necessary Jacobi-formula and cofactor identities; it does not assume the result it derives. Corollary 2 follows by applying Theorem 1 to constructed coordinate transformations, and the reverse-time Itô lemma is derived by two complementary routes—transform-then-reverse and reverse-then-transform—both from Anderson's external time-reversal formula and Itô's lemma. The GSSM objective is derived from score matching losses via substitution of Corollary 2 and integration by parts, with the original SSM recovered as the linear-projection special case. No parameter is fitted to reproduce the target results: the UCI experiments use fixed hyperparameters and benchmark against external SSM, while the chess example's scaling parameter w is an explicit post-hoc drift scaling, not a fitted prediction. The only substantive concerns—replacing the normalized loss (63) with the un-normalized (65) and the identifiability condition on the span of random gradients ∇v(x)—are correctness or assumption gaps, not circular reductions: they do not make the derivation equivalent to its inputs. There are also no load-bearing self-citations, as the paper cites only external prior work. Therefore, under the stated rules requiring a specific reduction to flag circularity, no circular step is present.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central theorem requires only standard calculus and the bijectivity assumption. The applications introduce two groups of hand-chosen hyperparameters (w and the quadratic variances), but no new physical or model entities. The paper does not invent new particles, mediators, or forces.

free parameters (2)
  • w (chess piece density scaling) = 0.8, 0.9, 1.0, 1.1 (Figure 5)
    Hand-chosen multiplicative factor on the reverse drift in the simplex diffusion example to control piece density; not learned and not tied to the central mathematical claims.
  • sigma_1, sigma_2, sigma_3 (GSSM random quadratic variance parameters) = sigma_3^2=1, sigma_2^2=1/sqrt(n), sigma_1^2=2/sqrt(n)
    Chosen a priori to set the scale of the random projection function v(x) in the GSSM experiments; affects estimator variance but is not fitted to improve benchmark scores.
assumptions (5)
  • domain assumption phi is bijective and twice continuously differentiable (Assumption 1)
    The main theorem requires a smooth invertible map; non-invertible or non-smooth transformations are excluded.
  • domain assumption Integrability and boundary conditions for score matching (Assumptions 2 and 3)
    Integration by parts in the GSSM derivation requires the boundary term pd(x)(grad v^T s_theta) grad v to vanish at infinity and finite moments.
  • standard math Anderson's reverse-time formula and Ito's lemma (standard stochastic calculus)
    Used in deriving the reverse-time SDE for the transformed process in Corollary 3.
  • standard math The identity (28) relating the log-determinant gradient to a divergence (stated in Remark 2)
    Used in the proof of Theorem 1; the identity as written is mis-stated, though the final theorem is correct.
  • domain assumption grad v(x) != 0 almost surely for the random projection in GSSM
    Corollary 2 and the GSSM loss divide by ||grad v||^2, so the projection direction must be nonzero; guaranteed for the quadratic choice when x != 0, which is an implicit assumption on the data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Score Change of Variables." pith.science (2026). https://pith.science/paper/Z5Y4YYH3

@misc{pith2026241207904,
  author       = {Pith},
  title        = {Pith review of: Score Change of Variables},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5Y4YYH3}},
  note         = {Machine review of arXiv:2412.07904}
}
abstract

We derive a general change of variables formula for score functions, showing that for a smooth, invertible transformation $\mathbf{y} = \phi(\mathbf{x})$, the transformed score function $\nabla_{\mathbf{y}} \log q(\mathbf{y})$ can be expressed directly in terms of $\nabla_{\mathbf{x}} \log p(\mathbf{x})$. Using this result, we develop two applications: First, we establish a reverse-time It\^o lemma for score-based diffusion models, allowing the use of $\nabla_{\mathbf{x}} \log p_t(\mathbf{x})$ to reverse an SDE in the transformed space without directly learning $\nabla_{\mathbf{y}} \log q_t(\mathbf{y})$. This approach enables training diffusion models in one space but sampling in another, effectively decoupling the forward and reverse processes. Second, we introduce generalized sliced score matching, extending traditional sliced score matching from linear projections to arbitrary smooth transformations. This provides greater flexibility in high-dimensional density estimation. We demonstrate these theoretical advances through applications to diffusion on the probability simplex and empirically compare our generalized score matching approach against traditional sliced score matching methods.

Figures

Figures reproduced from arXiv: 2412.07904 by the authors.

Figure 1
Figure 1. w = 0.8 [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 6
Figure 6. Exact Score Matching loss across three UCI datasets; lower loss is better. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Test log-likelihood for DKEF models trained with different loss functions. Higher values [PITH_FULL_IMAGE:figures/full_fig_p024_7.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Dale meets Langevin: A Multiplicative Denoising Diffusion Model

    cs.LG 2025-10 conditional novelty 6.0 of 10

    A geometric-Brownian-motion diffusion model with a multiplicative score-matching loss generates images through multiplicative updates that mirror exponentiated gradient descent.

Reference graph

Works this paper leans on

40 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [1]

    Reverse-time diffusion equation models

    Brian DO Anderson. Reverse-time diffusion equation models. Stochastic Processes and their Applications, 12(3):313–326, 1982

  2. [2]

    Uci machine learning repository, 2007

    Arthur Asuncion, David Newman, et al. Uci machine learning repository, 2007

  3. [3]

    Logistic-normal distributions: Some properties and uses

    Jhon Atchison and Sheng M Shen. Logistic-normal distributions: Some properties and uses. Biometrika, 67(2):261–272, 1980

  4. [4]

    Dirichlet diffusion score model for biological sequence generation

    Pavel Avdeyev, Chenlai Shi, Yuhao Tan, Kseniia Dudnyk, and Jian Zhou. Dirichlet diffusion score model for biological sequence generation. In International Conference on Machine Learning, pages 1276–1301. PMLR, 2023

  5. [5]

    Chess evaluations, March 2020

    Ronak Badhe. Chess evaluations, March 2020. Retrieved November 1, 2024, from https: //www.kaggle.com/datasets/ronakbadhe/chess-evaluations

  6. [6]

    Improving diffusion models for inverse problems using manifold constraints

    Hyungjin Chung, Byeongsu Sim, Dohoon Ryu, and Jong Chul Ye. Improving diffusion models for inverse problems using manifold constraints. Advances in Neural Information Processing Systems, 35:25683–25696, 2022

  7. [7]

    Riemannian score-based generative modelling

    Valentin De Bortoli, Emile Mathieu, Michael Hutchinson, James Thornton, Yee Whye Teh, and Arnaud Doucet. Riemannian score-based generative modelling. Advances in Neural Information Processing Systems, 35:2406–2422, 2022

  8. [8]

    Diffusion models beat gans on image synthesis

    Prafulla Dhariwal and Alexander Nichol. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34:8780–8794, 2021

Show all 40 references
  1. [9]

    Density estimation using real nvp, 2017

    Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. Density estimation using real nvp, 2017

  2. [10]

    Score-based generative modeling with critically-damped langevin diffusion

    Tim Dockhorn, Arash Vahdat, and Karsten Kreis. Score-based generative modeling with critically-damped langevin diffusion. arXiv preprint arXiv:2112.07068, 2021

  3. [11]

    Tweedie’s formula and selection bias

    Bradley Efron. Tweedie’s formula and selection bias. Journal of the American Statistical Association, 106(496):1602–1614, 2011

  4. [12]

    Partial differential equations, volume 19

    Lawrence C Evans. Partial differential equations, volume 19. American Mathematical Society, 2022. 10

  5. [13]

    Diffusion models for constrained domains, 2024

    Nic Fishman, Leo Klarner, Valentin De Bortoli, Emile Mathieu, and Michael Hutchinson. Diffusion models for constrained domains, 2024

  6. [14]

    Diffusion on the probability simplex, 2023

    Griffin Floto, Thorsteinn Jonsson, Mihai Nica, Scott Sanner, and Eric Zhengyu Zhu. Diffusion on the probability simplex, 2023

  7. [15]

    Real analysis: modern techniques and their applications, volume 40

    Gerald B Folland. Real analysis: modern techniques and their applications, volume 40. John Wiley & Sons, 1999

  8. [16]

    Denoising diffusion probabilistic models, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models, 2020

  9. [17]

    Cascaded diffusion models for high fidelity image generation

    Jonathan Ho, Chitwan Saharia, William Chan, David J Fleet, Mohammad Norouzi, and Tim Salimans. Cascaded diffusion models for high fidelity image generation. Journal of Machine Learning Research, 23(47):1–33, 2022

  10. [18]

    Estimation of non-normalized statistical models by score matching

    Aapo Hyvärinen and Peter Dayan. Estimation of non-normalized statistical models by score matching. Journal of Machine Learning Research, 6(4), 2005

  11. [19]

    Cocogen: Physically-consistent and conditioned score-based generative models for forward and inverse problems

    Christian Jacobsen, Yilin Zhuang, and Karthik Duraisamy. Cocogen: Physically-consistent and conditioned score-based generative models for forward and inverse problems. arXiv preprint arXiv:2312.10527, 2023

  12. [20]

    Brownian motion and stochastic calculus, volume 113

    Ioannis Karatzas and Steven Shreve. Brownian motion and stochastic calculus, volume 113. springer, 2014

  13. [21]

    A review of change of variable formulas for generative modeling, 2023

    Ullrich Köthe. A review of change of variable formulas for generative modeling, 2023

  14. [22]

    Reflected diffusion models, 2023

    Aaron Lou and Stefano Ermon. Reflected diffusion models, 2023

  15. [23]

    Generative fractional diffusion models

    Gabriel Nobis, Maximilian Springenberg, Marco Aversa, Michael Detzel, Rembert Daems, Roderick Murray-Smith, Shinichi Nakajima, Sebastian Lapuschkin, Stefano Ermon, Tolga Birdal, et al. Generative fractional diffusion models. arXiv preprint arXiv:2310.17638, 2023

  16. [24]

    Black box variational inference

    Rajesh Ranganath, Sean Gerrish, and David Blei. Black box variational inference. In Artificial intelligence and statistics, pages 814–822. PMLR, 2014

  17. [25]

    Variational inference with normalizing flows, 2016

    Danilo Jimenez Rezende and Shakir Mohamed. Variational inference with normalizing flows, 2016

  18. [26]

    High- resolution image synthesis with latent diffusion models

    Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. High- resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 10684–10695, 2022

  19. [27]

    U-net: Convolutional networks for biomedical image segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. U-net: Convolutional networks for biomedical image segmentation. In Medical image computing and computer-assisted intervention–MICCAI 2015: 18th international conference, Munich, Germany, October 5-9, 2015, proceedings, part ...

  20. [28]

    A spectral approach to gradient estimation for implicit distributions

    Jiaxin Shi, Shengyang Sun, and Jun Zhu. A spectral approach to gradient estimation for implicit distributions. In International Conference on Machine Learning, pages 4644–4653. PMLR, 2018

  21. [29]

    Sliced score matching: A scalable approach to density and score estimation, 2019

    Yang Song, Sahaj Garg, Jiaxin Shi, and Stefano Ermon. Sliced score matching: A scalable approach to density and score estimation, 2019

  22. [30]

    Yang Song and Diederik P. Kingma. How to train your energy-based models, 2021

  23. [31]

    Score-based generative modeling through stochastic differential equations

    Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020

  24. [32]

    A connection between score matching and denoising autoencoders

    Pascal Vincent. A connection between score matching and denoising autoencoders. Neural computation, 23(7):1661–1674, 2011

  25. [33]

    Learning deep kernels for exponential family densities

    Li Wenliang, Danica J Sutherland, Heiko Strathmann, and Arthur Gretton. Learning deep kernels for exponential family densities. In International Conference on Machine Learning, pages 6737–6746. PMLR, 2019. 11 A Proofs A.1 Proof of Theorem 1: Score Change of Variables in Rn Pro...

  26. [34]

    Transform-then-Reverse: We first transform the forward process using Itô’s lemma and then apply the time-reversal formula to the transformed process

  27. [35]

    Both approaches lead to the same result, demonstrating the path-wise equivalence: Reverse(ϕ(Xt, t)) ≡ ϕ(Reverse(Xt), t)

    Reverse-then-Transform: We first construct the reverse-time SDE of the original process and then transform it using Itô’s lemma. Both approaches lead to the same result, demonstrating the path-wise equivalence: Reverse(ϕ(Xt, t)) ≡ ϕ(Reverse(Xt), t). We also show the reverse-ti...

  28. [36]

    Epd [∥∇x log pd(X)∥2] < ∞

  29. [37]

    Assumption 3 (Boundary Conditions)

    For any v ∼ pv, Epd [∥∇v(X)∥2] < ∞ and ∇v(X) ̸= 0 almost surely. Assumption 3 (Boundary Conditions). For any v ∼ pv: pd(x)(∇v(x)⊤sθ(x))∇v(x) → 0 as ∥x∥ → ∞ (59) C.1 Score Matching in Transformed Space Consider a random vector X ∈ Rn with density pd(x). Let v : Rn → R be a rand...

  30. [38]

    Thus we require x ̸= 0

    for i < j – Independence of entries, except that off-diagonal entries Aij and Aji are dependent due to symmetry • Vector b: – b ∼ N(0, σ2 3I) – Independent from A D.1 Quadratic GSSM Loss The components needed for the GSSM loss are as follows: ∇v(x) = Ax + b (76) Hv(x) = A (77)...

  31. [39]

    Sample z ∼ N(0, 1), independently of ϵ

  32. [40]

    D.4 Additional Experimental Results Beyond the score matching loss presented in the main text, we also evaluated the models using test log-likelihood

    Compute Ax as: Ax = σ√ 2 (∥x∥ϵ + zx) . D.4 Additional Experimental Results Beyond the score matching loss presented in the main text, we also evaluated the models using test log-likelihood. These estimates were computed using Annealed Importance Sampling (AIS) with a proposal ...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.