REVIEW 5 major objections 5 minor 1 cited by
Integration Flow Models
T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash
Pith's one-line read Integration Flow is a generative framework that learns the integral of an ODE trajectory directly, so diffusion, rectified flow, and PFGM++ can generate in one network evaluation.
desk verdict One-step results worth a look, but the printed training loop is degenerate and Theorem 2 contradicts the paper's own Eq. (7); not ready for serious scrutiny. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The load-bearing identity is the antiderivative relation $\int_t^0 v(x_s, s)\,ds = V(x_0, 0) - V(x_t, t)$, rewritten as $x_0 = g(x_0, x_t, t) = x_t - G(x_0, x_t, t)$. A neural network $G_\theta$ replaces $G$, and the time-dependent coefficients $a_t$ and $b_t$ reparameterize the map to keep training stable when noise variance is large, as it is in VE diffusion and PFGM++. The iterative update $x_0^{(n+1)} = a_t x_t + b_t G_\theta(x_0^{(n)}, x_t, t)$ is what lets the estimate converge, and the Pseudo-Huber loss $d(x, y) = \sqrt{\|x-y\|_2^2 + c^2} - c$ is what the network is trained against.
What would settle it
Take an ODE-based generative model with a closed-form velocity, such as VE diffusion on a Gaussian mixture whose score is known exactly; if a trained Integration Flow's one-step output deviates from the exact ODE solution by more than the discretization error of a standard solver at one step, then the claim that it learns the integral without solver error would be falsified.
Extended reading notes
Core claim
Integration Flow defines the map $g_\theta(x_0, x_t, t) = a_t x_t + b_t G_\theta(x_0, x_t, t)$, where $G_\theta$ approximates the integrated reverse-time dynamics $G(x_0, x_t, t) = \int_t^0 v(x_s, s)\,ds$ from an intermediate state $x_t$ to the target $x_0$, with $x_0$ used as an explicit anchor. The paper argues that after training, $g_\theta(x_0^{(n)}, x_t, t)$ refines an estimate $x_0^{(n)}$ toward the true $x_0$, so that one pass of the network is enough to generate a sample. The authors claim that this single structure covers VE diffusion, rectified flow, and PFGM++, that it makes 1-rectified flow exactly straight without reflow, and that the anchor-state formulation is provably at least as accurate in mean squared error as conditioning on $x_t$ alone.
Load-bearing premise
The reported one-step results rest on the iterative training update $x_0^{(n+1)} = a_t x_t + b_t G_\theta(x_0^{(n)}, x_t, t)$ being a well-defined and effective procedure, but the paper specifies reusing an iterate from a previous, independently sampled data point and time rather than unrolling iterations for the same sample, so the training dynamic that produces the reported accuracy is left underspecified.
Editorial extensions
If this is right
- For the three models tested, one function evaluation replaces an ODE solve: VE diffusion reaches FID 2.86 on CIFAR-10 and 4.09 on ImageNet 64x64 at NFE 1.
- Rectified flow reaches competitive one-step quality without reflow, with FID 3.36 on CIFAR-10 and 4.35 on ImageNet, making the learned flow exactly straight in the 1-rectified case.
- PFGM++ becomes a one-step generator, with FID 2.91 on CIFAR-10 and 4.15 on ImageNet, which the authors state has not been shown before.
- The same architecture, loss, and training loop apply to all three models, with only the noise scheduler and the $a_t, b_t$ coefficients changing, so the framework unifies models that previously required separate treatment.
- A second network evaluation further improves quality, such as the CIFAR-10 VE FID dropping from 2.86 to 2.62, so the learned integral also serves as the initialization for cheap few-step refinement.
Reading between the lines
- If the antiderivative identity generalizes beyond the three implementations in the paper, the same $g_\theta$ construction should apply to any well-defined ODE-based generative model, including stochastic interpolants and flow matching variants, by setting $a_t$ and $b_t$ from the noise scheduler.
- The stability theorem's conditioning argument suggests that feeding the current estimate $x_0^{(n)}$ back into the network should make error shrink with each iteration; a direct test would be to measure whether the multistep FID improvement matches the contraction predicted by the bi-Lipschitz bound.
- The reported exact straightness of 1-rectified flow implies the training objective implicitly regularizes trajectory curvature; this could be tested by measuring the expected path curvature of learned trajectories against a baseline rectified flow trained with the usual matching loss.
- The paper notes extra training memory comes from carrying $x_0^{(n)}$; an off-GPU buffer would make the method cheaper to train, so the memory cost is an engineering constraint rather than a fundamental one.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper proposes Integration Flow, a framework that directly learns the integrated effect of an ODE-based generative trajectory rather than solving the ODE numerically. The method defines g(x0, xt, t) = xt - G(x0, xt, t), incorporating the target state x0 as an "anchor," and uses an iterative refinement x0^(n+1) = at xt + bt G_theta(x0^(n), xt, t). The authors provide three theorems (stability, non-intersection, optimality), derive specialized forms for VE diffusion, Rectified Flow, and PFGM++, and report one-step FIDs on CIFAR-10 and ImageNet 64x64 (e.g., 2.86, 3.36, 2.91 for VE, RF, PFGM++ on CIFAR-10). The paper claims to be the first unified structure for ODE-based generative models and the first to show exact straightness of 1-Rectified Flow without reflow.
Significance. If the empirical results were reproducible and the theory sound, the paper would make a valuable contribution: a single framework enabling one-step generation across several ODE-based generative models with competitive FIDs would be practically significant. The related work is reasonably comprehensive and the use of Pseudo-Huber loss and preconditioning is sensible. However, the manuscript does not include code, model checkpoints, or error bars, and the theoretical support contains serious internal inconsistencies. The central training algorithm as printed makes the iterative input statistically independent of the target, so the reported one-step results cannot be attributed to the described method. The paper's strengths are its clear organization and the breadth of empirical comparisons, but these do not compensate for the load-bearing flaws in the method description and theory.
major comments (5)
- [Section 3.3, Eq. (11); Algorithm 1] The training loop in Algorithm 1 (and Algorithms 3 and 5) carries x0^(n) across independent training steps. At step n, x0^(n) is the output produced for a previously sampled (x0, xt, t), and the current (x0, xt, t) is drawn independently. Hence x0^(n) carries no information about the current x0, and the Bayes-optimal predictor E[x0 | xt, x0^(n)] equals E[x0 | xt]; any dependence of g_theta on x0^(n) can only increase the loss. Theorem 1, which justifies the anchor state by comparing E[A|B,C] with E[A|B], does not apply because its C is uninformative under this procedure. At inference (Algorithm 2), iterations are run on the same xT, so x0^(n) becomes correlated with xT, a regime the training loop never creates. No inner loop unrolling iterations for the same sample is specified. As written, the training procedure forces the network to ignore x0^(n), and the reported one-step FIDs (Sections 5.2, Tables 2 and 3) are not reproducible from the description. Please provide a correct training algorithm (e.g., unrolling k steps per sample with backpropagation through the unrolled graph, or a clarified procedure) and re-run the experiments under that procedure.
- [Section 3.2, Eq. (7); Section 3.4, Theorem 2; Appendix B.2] Theorem 2 is internally inconsistent with Eq. (7). Equation (7) defines g(x0, xt, t) := x0, independently of xt for a fixed x0. Then for any two distinct states xt != yt with the same x0, the left-hand side of Eq. (13) is ||g(x0, xt, t) - g(x0, yt, t)|| = ||x0 - x0|| = 0, which violates the claimed lower bound e^{-Lt}||xt - yt|| > 0. The proof in Appendix B.2 treats g as the flow map of the reverse ODE, i.e., the solution at time 0 starting from xt, but the earlier definition identifies g with x0 itself. The bi-Lipschitz property and the non-intersection conclusion therefore do not follow from the stated definitions; Theorem 2 must be reformulated or removed.
- [Section 3.4, Theorem 1; Appendix B.1] Theorem 1 is a conditional-variance inequality: it states that E[(A - E[A|B,C])^2] <= E[(A - E[A|B])^2]. This is a standard fact, but the proof in Appendix B.1 asserts Var(A|B,C) <= Var(A|B) without proving it, and more importantly the theorem is used to justify the anchor state x0^(n). Under the training procedure in Algorithm 1, x0^(n) is not informative about x0, so the inequality is vacuous for the method as described. The theorem also assumes the estimators g'_theta and g_theta can represent the exact conditional expectations, which is not discussed. The theoretical justification of the anchor state is therefore not established.
- [Section 3.4, Theorem 3; Appendix B.3] The claim that Integration Flow is "optimal" for Flow Matching/Rectified Flow is not supported by the proof. The proof uses Jensen's inequality to show that the time-integrated objective is a lower bound on the flow-matching objective, i.e., min_v E ||∫(x1 - x0 - v(xt,t))dt||^2 <= min_v E ∫||(x1 - x0) - v(xt,t)||^2 dt. This inequality alone does not imply that the integrated approach achieves the best possible value; it only shows that the integrated objective is a relaxation. No argument is given that the minimum of the relaxed objective equals the minimum of the original objective, or that the learned G attains that minimum. The theorem's conclusion, "a direct one-step method (via G) achieves the best possible value," is therefore not proven. Additionally, the theorem's notation (x0 for the source, x1 for the target) conflicts with the rest of the paper, where x0 is the data sample.
- [Section 5; Figure 2] The empirical evaluation is not independently verifiable: no code, checkpoints, or random seeds are provided, and Tables 2 and 3 report FIDs without error bars. The claim that "the flow has already become exactly straight" for 1-Rectified Flow (Section 5.2, Figure 2) is not supported by any quantitative straightness metric or by a comparison of trajectory curvature before and after training. The reported one-step results also depend on the training/inference mismatch described above, so as they stand they do not validate the method. Please provide code/checkpoints or at least detailed experimental logs, and either provide a direct measurement of path straightness or temper the claim.
minor comments (5)
- [Appendix B.3] In Theorem 3 and its proof, the notation x0 and x1 is swapped relative to the main text, where x0 is the target data sample and xT (or z1) is the noise. This should be fixed to avoid confusion.
- [Algorithms 4 and 6] Both algorithms contain a duplicated initialization line: "initialize x0^(0) ~ N(0,I)" appears twice in Algorithm 4, and similar duplication occurs in Algorithm 6. One initialization suffices.
- [Figure 2] The caption says panels (a)-(b) are "regenerated from (Liu et al., 2022)" but it is unclear which panels are original and which are produced by the proposed method; add axis labels and a clear legend.
- [Table 2] The row for "1-Rectified flow" reports NFE=1, FID=378; this seems unusually high even for one-step rectified flow and may be a typographical error (e.g., 37.8 or 3.78). Please verify.
- [Appendix A.1, Eq. (32)] Equation (32) defines g_theta(x0, xt, t) = κ(σt) xt + 1/(1-κ(σt)) G_theta(x0, xt, t), but the text and Table 1 specify b_t = 1 - κ(σt), so the coefficient in front of G_theta appears inconsistent. Please check the derivation.
Circularity Check
Integration Flow's central derivation is definitional: g is set equal to x0 by Eq. (7), Theorem 3's 'optimality' uses G/t = x1-x0 by construction, and 'exact straightness' is the fitted regression target; the one-step results are a direct fit to x0.
-
self definitional
[Section 3.2, Equations (3)-(7)]
"This function encapsulates the total influence of the dynamics from an intermediate time t to the final time 0, which leads to the equation: x0 = xt−G(x0, xt,t ). (5) Next, we define the function: g(x0, xt,t ) := xt−G(x0, xt,t ). (6) Therefore, we have: x0 =g(x0, xt,t ). (7) Thus,g(x0, xt,t ) is the solution of the reversed time ODE from initial timet to final time 0, which encapsulates the cumulative effect of the reverse dynamics from the initial time t to the final time 0"
Equation (7) is not a consequence of the dynamics; it is the definition of g rearranged. Since G was defined as V(xt,t)-V(x0,0), Eq. (5) is just the fundamental theorem of calculus written as x0 = xt - G. Every later claim that g (or its network approximation) is 'the solution of the reversed time ODE' and that anchoring on x0 improves accuracy is already contained in the definition of g. The training loss d(gθ(x0^(n),xt,t),x0) then directly regresses the network to x0, so the one-step 'prediction' is the fitted target, not an independently derived forecast.
-
self definitional
[Appendix B.3, Theorem 3 proof, Equations (36)-(37)]
"= min_G E \left\| x1− x0− G(x1, xt,t)/t \right\|^2 // Integration Flow by Equation (36),(37) Therefore, Integration Flow models the transformation via a single integrated functionG, which is sufficient and effectively optimal for solving the Flow-Matching/Rectified-Flow objective."
In Eq. (36) the paper defines G(x0,zt,t) = t(z-x0), so in the theorem's notation G(x1,xt,t)/t = x1-x0. Substituting this identity makes the norm in the minimization identically zero for every G; the optimization is vacuous. Thus 'Integration Flow is optimal' is an artifact of the definition of G, not a property of a learned integrated flow.
1 more flagged steps
-
self definitional
[Section 5.2, Rectified Flow results paragraph]
"As shown in Figure 2, the flow has already become exactly straight for Integration Flow-based 1-rectified flow. This explains why Integration Flow-based 1-rectified flow can achieve good results as Reflow did."
For rectified flow, the Integration Flow map is set to gθ(x0,zt,t) = Gθ(x0,zt,t) (Eq. 19), and the target is x0 at every t along the linear interpolation zt = (1-t)x0 + tz. If the network is trained to output x0, the plotted flow from every zt to x0 is straight by construction; the figure confirms the training objective rather than discovering a straightening property. The claim to be 'the first to show exact straightness of 1-Rectified Flow without reflow' therefore reduces to the definition of the regression target.
full rationale
The central theoretical contributions of the paper reduce by construction. Equation (7) defines g so that x0 = g(x0,xt,t) is an identity, and Eq. (10) states that the network is trained so that gθ(x0,xt,t) ≈ x0. Theorem 3's 'optimality' proof substitutes G/t = x1-x0 from Eq. (36), making the minimized quantity identically zero. The exact-straightness claim for 1-Rectified Flow is likewise the training objective restated as a geometric finding. These are genuine cases of self-definitional circularity rather than independent derivation. Separately, Section 3.4 Theorem 1 proves an inequality for conditional expectations and then declares gθ 'analogous to' E[x0|xt,x0^(n)]; no argument identifies the trained network with that conditional expectation, and in Algorithms 1, 3, and 5 the iterate x0^(n) is carried from a previous, independently sampled training point, so the conditioning variable is uninformative. The paper also acknowledges in Section 6 that storing x0^(n) requires buffer or disk and says 'We will fix this in our future work.' These are correctness and reproducibility concerns rather than additional circularity, and they do not change the verdict: the one-step FID results, while empirically reported, are fits of a network to the target x0, and the theoretical claims that are supposed to justify them are definitional. Score 7 reflects that the central claims reduce by construction, with the empirical evaluations providing independent but not theory-supporting content.
Assumptions & free parameters
free parameters (5)
- Pseudo-Huber loss constant c (VE) =
0.00016
- Pseudo-Huber loss constant c (Rectified Flow) =
0.0001
- Pseudo-Huber loss constant c (PFGM++) =
0.00016
- Preconditioning coefficients a_t, b_t =
a_t=sigma_min/sigma_t, b_t=1-sigma_min/sigma_t for VE and PFGM++; a_t=0, b_t=1 for Rectified Flow
- sigma_min =
0.01
assumptions (5)
- domain assumption The reverse-time ODE (Eq. 1) has a unique solution and the velocity field v is Lipschitz continuous.
- domain assumption A neural network can be trained to essentially equal the true integrated map g(x0, xt, t), with sufficient capacity and optimization.
- ad hoc to paper The iterative estimator gtheta can be treated as the conditional expectation E[x0 | xt, x0^(n)].
- standard math The antiderivative V exists and the fundamental theorem of calculus applies to the reverse-time trajectory.
- domain assumption Solving the reverse-time ODE yields the target data distribution.
Cite this review
Pith. "Pith review of Integration Flow Models." pith.science (2026). https://pith.science/paper/V77NVS3E
@misc{pith2026250420179,
author = {Pith},
title = {Pith review of: Integration Flow Models},
year = {2026},
howpublished = {\url{https://pith.science/paper/V77NVS3E}},
note = {Machine review of arXiv:2504.20179}
}
abstract
Ordinary differential equation (ODE) based generative models have emerged as a powerful approach for producing high-quality samples in many applications. However, the ODE-based methods either suffer the discretization error of numerical solvers of ODE, which restricts the quality of samples when only a few NFEs are used, or struggle with training instability. In this paper, we proposed Integration Flow, which directly learns the integral of ODE-based trajectory paths without solving the ODE functions. Moreover, Integration Flow explicitly incorporates the target state $\mathbf{x}_0$ as the anchor state in guiding the reverse-time dynamics. We have theoretically proven this can contribute to both stability and accuracy. To the best of our knowledge, Integration Flow is the first model with a unified structure to estimate ODE-based generative models and the first to show the exact straightness of 1-Rectified Flow without reflow. Through theoretical analysis and empirical evaluations, we show that Integration Flows achieve improved performance when it is applied to existing ODE-based models, such as diffusion models, Rectified Flows, and PFGM++. Specifically, Integration Flow achieves one-step generation on CIFAR10 with FIDs of 2.86 for the Variance Exploding (VE) diffusion model, 3.36 for rectified flow without reflow, and 2.91 for PFGM++; and on ImageNet with FIDs of 4.09 for VE diffusion model, 4.35 for rectified flow without reflow and 4.15 for PFGM++.
Figures
Forward citations
Cited by 1 Pith paper
-
Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling
Modular MeanFlow interpolates between stop-gradient and full-gradient training for one-step generative models via a warmup schedule, reporting a CIFAR-10 FID of 3.41.
Reference graph
Works this paper leans on
-
[1]
Albergo, M. S. and Vanden-Eijnden, E. Building normal- izing flows with stochastic interpolants. arXiv preprint arXiv:2209.15571,
-
[8]
Multistep con- sistency models
Heek, J., Hoogeboom, E., and Salimans, T. Multistep con- sistency models. arXiv preprint arXiv:2403.06807,
-
[10]
Kim, D., Kim, Y ., Kwon, S. J., Kang, W., and Moon, I.- C. Refining generative process with discriminator guid- ance in score-based diffusion models. arXiv preprint arXiv:2211.17091,
-
[11]
9 Submission for ICML 2025 Krizhevsky, A., Hinton, G., et al
URL https://openreview.net/forum? id=a-xFK8Ymz5J. 9 Submission for ICML 2025 Krizhevsky, A., Hinton, G., et al. Learning multiple layers of features from tiny images
work page 2025
-
[14]
Flow straight and fast: Learning to generate and transfer data with rectified flow
Liu, X., Gong, C., and Liu, Q. Flow straight and fast: Learning to generate and transfer data with rectified flow. arXiv preprint arXiv:2209.03003,
-
[15]
Luhman, E. and Luhman, T. Knowledge distillation in iterative generative models for improved sampling speed. arXiv preprint arXiv:2101.02388,
-
[16]
Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models
Luo, W., Hu, T., Zhang, S., Sun, J., Li, Z., and Zhang, Z. Diff-instruct: A universal approach for transferring knowledge from pre-trained diffusion models. arXiv preprint arXiv:2305.18455,
-
[17]
Song, J., Meng, C., and Ermon, S
URL https:// arxiv.org/abs/2301.09515. Song, J., Meng, C., and Ermon, S. Denoising diffusion im- plicit models. arXiv preprint arXiv:2010.02502, 2020a. Song, Y . and Dhariwal, P. Improved techniques for training consistency models. arXiv preprint arXiv:2310.14189 ,
arXiv 2010
Show all 25 references
-
[18]
P., Kumar, A., Er- mon, S., and Poole, B
Song, Y ., Sohl-Dickstein, J., Kingma, D. P., Kumar, A., Er- mon, S., and Poole, B. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456, 2020b. Song, Y ., Dhariwal, P., Chen, M., and Sutskever, I. Consis- tency models. arX...
2011 arXiv
-
[19]
Dropout: a simple way to prevent neural networks from overfitting
Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I., and Salakhutdinov, R. Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 15(1):1929–1958,
1929
-
[20]
Consistency flow matching: Defining straight flows with velocity consis- tency
Yang, L., Zhang, Z., Zhang, Z., Liu, X., Xu, M., Zhang, W., Meng, C., Ermon, S., and Cui, B. Consistency flow matching: Defining straight flows with velocity consis- tency. arXiv preprint arXiv:2407.02398,
-
[21]
Directly denoising diffu- sion model
Zhang, D., Wang, J., and Luo, F. Directly denoising diffu- sion model. arXiv preprint arXiv:2405.13540,
-
[22]
and Chen, Y
Zhang, Q. and Chen, Y . Fast sampling of diffusion models with exponential integrator. arXiv preprint arXiv:2204.13902,
-
[23]
Unipc: A unified predictor-corrector framework for fast sampling of diffusion models
Zhao, W., Bai, L., Rao, Y ., Zhou, J., and Lu, J. Unipc: A unified predictor-corrector framework for fast sampling of diffusion models. arXiv preprint arXiv:2302.04867 ,
-
[24]
Derivation of Integration Flow Algorithms A.1
10 Submission for ICML 2025 A. Derivation of Integration Flow Algorithms A.1. Integration Flow for VE diffusion models The PF-ODE of VE diffusion models is formulated as: dxt dt =− 1 2 dσ2 t dt∇xt logpt (xt) (27) We do the reversed time integration on both sides over the inter...
2025
-
[25]
showed thatdx/dr =dx/dσ,whereσ changes with time. We modify the (39) as: dx dt = dx dr dσt dt = E(˜x)x E(˜x)r dσt dt (40) We do the reversed time integration on both sides with respect tot over the interval [0,t ], leading to: Z 0 t dx dtdt = Z 0 t E(˜x)x E(˜x)r dσt dt dt, (41...
2025
-
[1997]
doi: 10.1109/83.551699. Chen, R. T., Rubanova, Y ., Bettencourt, J., and Duvenaud, D. K. Neural ordinary differential equations. Advances in neural information processing systems , 31,
-
[2009]
Improving the training of rectified flows
Lee, S., Lin, Z., and Fanti, G. Improving the training of rectified flows. arXiv preprint arXiv:2405.20320,
-
[2018]
Imagenet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pp. 248–255. Ieee,
2009
-
[2019]
Boot: Data-free distillation of denoising diffusion models with bootstrapping
Gu, J., Zhai, S., Zhang, Y ., Liu, L., and Susskind, J. Boot: Data-free distillation of denoising diffusion models with bootstrapping. arXiv preprint arXiv:2306.05544,
-
[2020]
Scalable adaptive computation for iterative generation
Jabri, A., Fleet, D., and Chen, T. Scalable adaptive computation for iterative generation. arXiv preprint arXiv:2212.11972,
-
[2021]
Anode: Uncondi- tionally accurate memory-efficient gradients for neural odes
Gholami, A., Keutzer, K., and Biros, G. Anode: Uncondi- tionally accurate memory-efficient gradients for neural odes. arXiv preprint arXiv:1902.10298,
1902 arXiv
-
[2022]
A., Zhai, S., Hu, S., Zheng, D., Talbot, W., and Gu, E
Berthelot, D., Autef, A., Lin, J., Yap, D. A., Zhai, S., Hu, S., Zheng, D., Talbot, W., and Gu, E. Tract: Denoising diffusion models with transitive closure time-distillation. arXiv preprint arXiv:2303.04248,
-
[2023]
Large scale gan training for high fidelity natural image synthesis
Brock, A., Donahue, J., and Simonyan, K. Large scale gan training for high fidelity natural image synthesis. arXiv preprint arXiv:1809.11096,
-
[2024]
T., Ben-Hamu, H., Nickel, M., and Le, M
Lipman, Y ., Chen, R. T., Ben-Hamu, H., Nickel, M., and Le, M. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747,
Reviewed August 16, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.