Pith. sign in

REVIEW 5 major objections 5 minor 48 references

Graph Guided Diffusion: Unified Guidance for Conditional Graph Generation

T0 review · 5 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper proposes GGDiff, which treats conditional graph diffusion as a stochastic optimal control problem and approximates the optimal control with one-step denoising plus zero-order search, enabling guidance of pre-trained models by…

desk verdict A useful, honestly-reported application of SOC guidance to graph diffusion; the best-of-N zero-order variant is a practical contribution, but the unification claim and superiority claims need a sharper derivation, code, and stronger baselines before they fully hold. read the letter →

arxiv 2505.19685 v1 pith:ACECMRZH submitted 2025-05-26 cs.LG

Victor M. Tenorio , Nicolas Zilberstein , Santiago Segarra , Antonio G. Marques This is my paper · ORCID
classification cs.LG
keywords graphgenerationdiffusionmodelsconditionalstochasticoptimalcontrolzero-orderoptimizationnon-differentiablerewardsguidancefairnessconstraints
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

The paper proposes GGDiff, a framework for conditional graph generation that works with arbitrary reward functions, including non-differentiable ones. Its central claim is that guiding a pre-trained continuous graph diffusion model can be treated as a stochastic optimal control problem, and that the resulting optimal control can be made tractable by evaluating the reward on a one-step denoised estimate of the graph plus randomized-direction search. If correct, GGDiff would let users impose structural constraints, fairness criteria, or partial-observation conditions on generated graphs without retraining the model, computing gradients of the reward, or designing projection operators. Experiments on motif constraints, fair graph generation, and link prediction report better reward alignment than a projection-based baseline while preserving the distribution of the underlying graph family.

What carries the argument

The load-bearing object is the optimal control law $U^*(G_t^C,t) = -g(t)\nabla_{G_t^C}\log \mathbb{E}_{\mathrm{pre}}[\exp(-r(G_0^C)/\lambda)\mid G_t^C]$ from the Feynman-Kac solution of the linearized Hamilton-Jacobi-Bellman equation. Since the expectation is intractable, the paper applies a greedy approximation: $p(G_0^C\mid G_t^C)$ is treated as a Dirac delta at the one-step MMSE denoiser $\hat{G}_0^C(G_t^C)$ from Tweedie's formula, turning the control into the reward gradient $\frac{g(t)}{\lambda}\nabla_{\hat{G}_t^C}r(\hat{G}_0^C(\hat{G}_t^C))$. For non-differentiable rewards, the gradient is replaced by a general zero-order estimator $\hat{\nabla}r(\hat{G}_t^C):=\mathbb{E}_{U_t\sim D}[w(U_t)r(\hat{G}_0^C(\hat{G}_t^C+\mu U_t))U_t]$, instantiated as one-point, two-point, best-of-N, and multi-point versions. The one-step denoiser is what connects the reward (defined on clean graphs) to the noisy state, and all guidance rules inherit their quality from it.

What would settle it

Run GGDiff on the ego-small dataset with a deliberately corrupted one-step denoiser active only in the first half of the reverse process: if reward alignment and validity stay unchanged while the corruption is active, the Dirac-delta approximation is not load-bearing; if they degrade sharply, the central claim depends on the denoiser being accurate at coarse noise levels.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that conditional graph generation should be posed as a stochastic control problem: the sampler follows a controlled diffusion whose drift contains an extra control term, and the optimal control is given by the Feynman-Kac formula as the gradient of a log-expectation of the exponentiated reward. Because that expectation is intractable, the paper replaces the conditional distribution of the clean graph with a Dirac delta at the one-step Tweedie denoiser, yielding a greedy control proportional to the gradient of the reward at the denoised estimate; for non-differentiable rewards it replaces the gradient with zero-order estimators—one-point, two-point, best-of-N, and multi-point random search—that evaluate the reward along random directions. The paper argues this single formulation unifies gradient-based guidance, control-based guidance, and gradient-free optimization, and that it enables zero-shot guidance of pre-trained diffusion models under both differentiable and black-box rewards.

Load-bearing premise

The guidance direction is computed by treating the one-step denoised graph as if it were the true final graph (a Dirac-delta approximation), so if that denoiser is inaccurate—especially early in the reverse process—the reward signal is misaligned and the steering can push samples off the data manifold.

Editorial extensions

If this is right

  • A user can condition a pre-trained graph diffusion model on a black-box reward with no retraining, no reward gradient, and no closed-form projection operator.
  • Differentiable rewards automatically recover the standard gradient-guidance rule used in diffusion posterior sampling, so the framework contains existing gradient guidance as a special case.
  • Non-differentiable constraints become expressible directly—for example, quantized adjacency matrices or one-sided penalties—expanding the space of graph conditions that can be targeted.
  • Empirically, the framework reports higher constraint-satisfaction rates and better distribution preservation than the projection-based baseline on degree, edge-count, and triangle constraints, and it handles fairness and partial-observation tasks that the baseline cannot.
  • Because all guidance is computed through reward evaluations on the one-step denoiser, the method's computational cost is controlled by the number of reward queries per diffusion step, with best-of-N giving low variance in discrete settings.

Reading between the lines

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

  • Because the greedy Dirac-delta approximation is known by the authors to degrade away from $t=0$, a natural extension is an adaptive schedule that throttles guidance strength or uses multi-step denoisers early in the reverse process; the paper does not test this.
  • The same control-plus-zero-order recipe is not graph-specific once a denoiser and a reward are available, so GGDiff's machinery could plausibly transfer to other structured outputs such as point clouds, sets, or trajectories.
  • A direct comparison between GGDiff's greedy control and an exact Monte Carlo evaluation of the Feynman-Kac expectation on small graphs would isolate how much performance comes from the control formulation versus from the one-step approximation; the paper does not run this comparison.
  • The best-of-N estimator's bias toward high-reward directions suggests GGDiff may trade away some modes of the prior when rewards are multimodal; testing diversity under multi-peaked rewards would clarify that trade-off.
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

5 major / 5 minor

Summary. The paper proposes GGDiff, a framework for conditioning pre-trained continuous graph diffusion models on arbitrary reward functions. Conditional generation is formulated as a stochastic optimal control (SOC) problem, and the intractable optimal control of Eq. (4) is approximated by a greedy scheme built on the one-step denoiser: differentiable rewards use the gradient of the reward at the denoised sample (Eq. (6)), while non-differentiable rewards use one of three zero-order estimators (Eqs. (9)-(12)). The method is evaluated on structural constraints (maximum degree, edge count, triangle count, star graphs), fair graph generation, and incomplete graph generation, with comparisons to PRODIGY on the first set of tasks.

Significance. If the approximation in Eq. (5) is reliable and the loss-selection protocol is controlled, GGDiff would be a practically useful contribution: it brings gradient-free guidance to graph diffusion and unifies several known guidance heuristics under one control-theoretic umbrella. The paper provides a standard SOC derivation (Appendix A), releases code, and covers a broader set of constraint types than PRODIGY, including fairness and link prediction. At the same time, the central Dirac-delta approximation is not validated at coarse noise levels, the zero-order estimators are not formally connected to the SOC solution, and several empirical claims are stronger than the reported numbers support. These issues are fixable but require substantive additional analysis.

major comments (5)
  1. [Section 3.2, Eq. (5)] The Dirac-delta approximation p(G0^C | \hat G_t^C) = δ(G0^C - \hat G0^C(\hat G_t^C)) is the unique bridge between the exact SOC control in Eq. (4) and every implemented guidance rule, including the zero-order estimators of Eqs. (9)-(12), which all evaluate the reward at the one-step denoiser. At coarse noise levels the posterior p(G0 | G_t) is broad and the one-step denoiser is known to be inaccurate, so the control direction can be misaligned and push samples off the data manifold. The paper's only safeguard is the remark in Section 3.2 that approximation errors diminish as t approaches 0; no error bound, no ablation of the guidance starting time, and no comparison with a multi-step denoising estimate are provided. Because this assumption is inherited by all GGDiff variants, the claimed general balance of reward alignment and fidelity is not established.
  2. [Section 3.2.2, Eqs. (7)-(12)] The zero-order guidance is introduced as the argmax in Eq. (7) over directions U_t of r(\hat G0^C(\hat G_t^C + μ U_t)), but no derivation connects this to the optimal control in Eq. (4) or to the gradient approximation in Eq. (6). In particular, the best-of-N rule in Eq. (11) selects a direction that maximizes the reward after perturbation without any gradient magnitude, expectation over the uncontrolled process, or path-integral interpretation. The text acknowledges these are 'pseudo-gradients,' but the paper's central claim of unifying gradient-based, control-based, and zero-order guidance under a single SOC formulation requires at least a formal statement of the sense in which Eq. (7) approximates Eq. (4). Without this, the theoretical contribution is a collection of heuristics rather than a unified framework.
  3. [Appendix E.3.1 (and Table 1)] The manuscript states that the loss function for each constraint is 'empirically selected from a pool of possibilities based on which yields the best performance.' Since this selection is made after seeing the evaluation metrics (ValC and ΔMMD) on the test sets, the results in Table 1 are at risk of selection on the test set, and they do not support the zero-shot, plug-and-play claim for arbitrary rewards. The authors should either specify a fixed selection protocol using a validation split and report the number and identity of candidate losses, or reframe the method as requiring per-task tuning. As written, the reported ValC/ΔMMD numbers conflate the algorithm with its tuning.
  4. [Table 1 and Section 4.1] The abstract and Section 4 claim that GGDiff achieves 'superior alignment with target rewards while maintaining diversity and fidelity,' but Table 1 reports negative ΔMMD values for most GGDiff cells, e.g., GGDiff-C on Community Small Max Degree (−0.73 with ValC 1.00), GGDiff-G on Enzymes Max Degree (−0.37), and GGDiff-Z on Ego Small Triangle Count (−0.07). Since the unconstrained baseline is defined as 0.00, these values indicate that the generated graphs deviate more from the original data distribution than the unconstrained model, and in several cells PRODIGY has a better ΔMMD (e.g., PRODIGY 0.07 vs. GGDiff-G −0.37 on Enzymes Max Degree). The paper needs to acknowledge this trade-off explicitly and report combined statistics or a Pareto-style evaluation; the fidelity claim is contradicted by the paper's own headline table.
  5. [Sections 4.2 and 4.3] The fairness and incomplete-graph experiments compare only against the unconstrained model and not against any prior conditional graph generation baseline. Since the introduction states that GGDiff 'beats current state-of-the-art architectures' and achieves 'superior performance and flexibility over existing methods,' the absence of competitive baselines in two of the three experimental areas leaves the state-of-the-art claim unsupported outside the PRODIGY comparison in Section 4.1. The authors should add at least one baseline per task (e.g., classifier guidance on DiGress or an adapted PRODIGY variant where possible) or restrict the claims to what the experiments actually compare.
minor comments (5)
  1. [Appendix A, Eq. (16)] The boundary condition below Eq. (14) states V0(G0) = r(G0), but Eq. (13) defines the value function as an expectation involving −r(G0), which implies V0(G0) = −r(G0). This sign inconsistency should be corrected in the derivation.
  2. [Section 3.2.2, after Eq. (9)] The text says the variance of the one-point estimator 'explodes as µ increases to 0'; this should read 'as µ decreases to 0' or 'as µ → 0'.
  3. [Eq. (11)] The notation U_t^{(i)} = argmax_{...} r(...) · U_t is not well-formed because the argmax returns an index, not a direction. Define i* = argmax_i r(\hat G0^C(\hat G_t^C + μ U_t^{(i)})) and then set U_t = k U_t^{(i*)}.
  4. [Algorithm 1, lines 5-6] Algorithm 1 computes U_t = ∇_{\hat G_t} r(\hat G0^C(\hat G_t^C)), but the derivation in Eq. (6) includes the factor g(t)/λ. The implementation should state how the step size k, g(t), and λ are combined, or the algorithm should match the equation.
  5. [Appendix E] The values of the key hyperparameters N, k, μ, λ, and the guidance schedule are not reported anywhere in the main text or appendices; since the method's behavior is sensitive to these quantities, they should be listed per dataset and task to make the experiments reproducible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the SOC derivation is self-contained, the greedy approximation is explicit, and the empirical self-citation is not load-bearing.

full rationale

The central derivation chain is independent of its conclusions. The optimal control formula in Eq. (4) follows from the Feynman-Kac solution of the linearized HJB equation derived in Appendix A, which is an external mathematical result. The greedy approximation in Eq. (5) is an explicitly stated modeling assumption -- that the posterior p(G0|Gt) is a Dirac delta at the one-step denoiser -- and Eq. (6) is obtained by substituting that approximation into Eq. (4), with the paper clearly labeling it an approximation that improves as t approaches 0. The non-differentiable estimators in Eqs. (8)-(12) are standard zeroth-order constructions (one-point, two-point, best-of-N, multi-point); defining a general estimator and instantiating it is not circular. The only mild self-reference is in Section 4.2 and Appendix E.3.2, where the fairness loss and metrics come from Navarro et al. [21], a paper co-authored by two of the present authors. However, the fairness definition is an externally published criterion, and the GGDiff framework does not depend on it for its derivation, so the self-citation is not load-bearing. The empirical loss selection noted in Appendix E.3.1 is a potential overfitting or selection-bias concern, but the loss is an input reward rather than a predicted output, so it does not make the derivation circular. Overall, no claimed prediction or first-principles result reduces by construction to its inputs.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The method relies on standard optimal-control mathematics and Tweedie denoising, plus two paper-specific assumptions: the Dirac-delta posterior approximation and the effectiveness of random perturbations as surrogate guidance. No fundamentally new physical or structural entities are introduced. The main uncharged load is the post-hoc selection of loss functions and the unreported hyperparameters mu, k, N, and lambda.

free parameters (6)
  • smoothing parameter mu
    Controls the perturbation scale in the zero-order estimators, used in Eqs. (9)-(12); no value is reported in the paper.
  • step size k
    Scales the control update in Algorithm 1, line 18; no value is reported.
  • number of candidate directions N
    Number of sampled directions in the best-of-N and multi-point estimators; no value is reported.
  • regularization weight lambda
    Appears in the SOC objective (3) and scales the guidance in Eq. (6); the paper says 1/lambda is a scaling factor but does not report its value.
  • loss function per constraint = one per constraint, chosen post hoc
    Appendix E.3.1 states the loss function was empirically selected from a pool based on which yields the best performance, so results are conditional on this selection.
  • SBM validity threshold factor = 8
    Section 4.2 chooses a factor of 8 so that 95% of test graphs are valid SBMs, an ad hoc calibration of the evaluation criterion.
assumptions (4)
  • standard math Feynman-Kac formula gives the optimal control as U* = -g(t) times the gradient of the log-expectation of exp(-r(G0)/lambda) conditional on Gt (Eq. 4).
    Invoked in Section 3.1 and derived in Appendix A through the linearized HJB equation.
  • standard math Tweedie's formula yields the MMSE denoiser E[G0|Gt] = Ghat_0(Gt).
    Used in Section 2.2 and Section 3.2.1 to evaluate rewards at a one-step denoised estimate.
  • ad hoc to paper The posterior p(G0|Ghat_t) is concentrated at the one-step denoiser Ghat_0(Ghat_t).
    Eq. (5) replaces the conditional expectation with a Dirac delta at the denoised estimate; this greedy approximation underpins both the differentiable and zero-order control rules.
  • domain assumption Randomly perturbing the one-step denoised graph and re-evaluating the reward gives a useful surrogate control direction even when the reward is non-differentiable.
    Section 3.2.2 adopts this as the basis for all zero-order estimators; the paper acknowledges the direction is a surrogate, not an unbiased gradient.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Graph Guided Diffusion: Unified Guidance for Conditional Graph Generation." pith.science (2026). https://pith.science/paper/ACECMRZH

@misc{pith2026250519685,
  author       = {Pith},
  title        = {Pith review of: Graph Guided Diffusion: Unified Guidance for Conditional Graph Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ACECMRZH}},
  note         = {Machine review of arXiv:2505.19685}
}
read the original abstract

Diffusion models have emerged as powerful generative models for graph generation, yet their use for conditional graph generation remains a fundamental challenge. In particular, guiding diffusion models on graphs under arbitrary reward signals is difficult: gradient-based methods, while powerful, are often unsuitable due to the discrete and combinatorial nature of graphs, and non-differentiable rewards further complicate gradient-based guidance. We propose Graph Guided Diffusion (GGDiff), a novel guidance framework that interprets conditional diffusion on graphs as a stochastic control problem to address this challenge. GGDiff unifies multiple guidance strategies, including gradient-based guidance (for differentiable rewards), control-based guidance (using control signals from forward reward evaluations), and zero-order approximations (bridging gradient-based and gradient-free optimization). This comprehensive, plug-and-play framework enables zero-shot guidance of pre-trained diffusion models under both differentiable and non-differentiable reward functions, adapting well-established guidance techniques to graph generation--a direction largely unexplored. Our formulation balances computational efficiency, reward alignment, and sample quality, enabling practical conditional generation across diverse reward types. We demonstrate the efficacy of GGDiff in various tasks, including constraints on graph motifs, fairness, and link prediction, achieving superior alignment with target rewards while maintaining diversity and fidelity.

Figures

Figures reproduced from arXiv: 2505.19685 by the authors.

Figure 1
Figure 1. Illustration of GGDiff, a method that guides the generation of graphs to satisfy a set of constraints (in this case, the constraint is fairness). The guidance Ut is a local direction obtained via SOC, and approximated using ZO techniques, like the multi-point estimate shown here. these approaches require differentiable constraints, which limits their applicability in more complex graph generation tasks where constra… view at source ↗
Figure 2
Figure 2. Samples for the force stars constraint in the Ego small dataset. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Samples from the fair graph generation experiment. [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Samples generated for the incomplete graph generation experiment. The graph on the left [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

48 extracted references · 43 canonical work pages

  1. [1]

    D., Ho, J., Tarlow, D., and Van Den Berg, R

    Austin, J., Johnson, D. D., Ho, J., Tarlow, D., and Van Den Berg, R. (2021). Structured denoising diffusion models in discrete state-spaces. Advances in Neural Inf. Process. Syst. (NeurIPS) , 34:17981–17993

    work page 2021

  2. [2]

    S., Cao, L., Choromanski, K., and Scheinberg, K

    Berahas, A. S., Cao, L., Choromanski, K., and Scheinberg, K. (2022). A theoretical and empirical comparison of gradient approximations in derivative-free optimization. Foundations of Comp. Math., 22(2):507–560

    work page 2022

  3. [3]

    Boyd, S. P. and Vandenberghe, L. (2004). Convex optimization. Cambridge university press

    work page 2004

  4. [4]

    Chen, X., He, J., Han, X., and Liu, L.-P. (2023). Efficient and degree-guided graph generation via discrete diffusion modeling. Intl. Conf. on Machine Learning (ICML)

    work page 2023

  5. [5]

    T., Klasky, M

    Chung, H., Kim, J., Mccann, M. T., Klasky, M. L., and Ye, J. C. (2022). Diffusion posterior sampling for general noisy inverse problems. In Intl. Conf. Learn. Repr. (ICLR)

    work page 2022

  6. [6]

    Evans, L. C. (2022). Partial differential equations, volume 19. American Mathematical Society

    work page 2022

  7. [7]

    Grover, A., Zweig, A., and Ermon, S. (2019). Graphite: Iterative generative modeling of graphs. In Intl. Conf. on Machine Learning (ICML), pages 2434–2444. PMLR

    work page 2019

  8. [8]

    Ho, J., Jain, A., and Abbeel, P. (2020). Denoising diffusion probabilistic models. Advances in Neural Inf. Process. Syst. (NeurIPS), 33:6840–6851

    work page 2020

Show all 48 references
  1. [9]

    and Salimans, T

    Ho, J. and Salimans, T. (2021). Classifier-free diffusion guidance. In NeurIPS 2021 Workshop on Deep Generative Models and Downstream Applications

    2021

  2. [10]

    G., Vignac, C., and Welling, M

    Hoogeboom, E., Satorras, V . G., Vignac, C., and Welling, M. (2022). Equivariant diffusion for molecule generation in 3d. In Intl. Conf. on Machine Learning (ICML), pages 8867–8887. PMLR

    2022

  3. [11]

    S., Gururani, S., Oore, S., and Yue, Y

    Huang, Y ., Ghatare, A., Liu, Y ., Hu, Z., Zhang, Q., Sastry, C. S., Gururani, S., Oore, S., and Yue, Y . (2024). Symbolic music generation with non-differentiable rule guided diffusion.arXiv preprint arXiv:2402.14285

    2024 arXiv

  4. [12]

    Jo, J., Lee, S., and Hwang, S. J. (2022). Score-based generative modeling of graphs via the system of stochastic differential equations. In Intl. Conf. on Machine Learning (ICML), pages 10362–10383. PMLR

    2022

  5. [13]

    Larson, J., Menickelly, M., and Wild, S. M. (2019). Derivative-free optimization methods. Acta Numerica, 28:287–404. 10

    2019

  6. [14]

    Li, X., Zhao, Y ., Wang, C., Scalia, G., Eraslan, G., Nair, S., Biancalani, T., Ji, S., Regev, A., Levine, S., et al. (2024). Derivative-free guidance in continuous and discrete diffusion models with soft value-based decoding. arXiv preprint arXiv:2408.08252

    2024 arXiv

  7. [15]

    O., and Varshney, P

    Liu, S., Chen, P.-Y ., Kailkhura, B., Zhang, G., Hero III, A. O., and Varshney, P. K. (2020). A primer on zeroth-order optimization in signal processing and machine learning: Principals, recent advances, and applications. IEEE Signal Process. Mag., 37(5):43–54

    2020

  8. [16]

    Liu, S., Kailkhura, B., Chen, P.-Y ., Ting, P., Chang, S., and Amini, L. (2018). Zeroth-order stochastic variance reduction for nonconvex optimization. Advances in Neural Inf. Process. Syst. (NeurIPS), 31

    2018

  9. [17]

    Luo, T., Mo, Z., and Pan, S. J. (2023). Fast graph generation via spectral diffusion. IEEE Trans. on Patt. Analysis and Machine Int., 46(5):3496–3508

    2023

  10. [18]

    Madeira, M., Vignac, C., Thanou, D., and Frossard, P. (2024). Generative modelling of structurally constrained graphs. Advances in Neural Inf. Process. Syst. (NeurIPS), 37:137218– 137262

    2024

  11. [19]

    Mardani, M., Song, J., Kautz, J., and Vahdat, A. (2024). A variational perspective on solving inverse problems with diffusion models. Intl. Conf. Learn. Repr. (ICLR)

    2024

  12. [20]

    Minello, G., Bicciato, A., Rossi, L., Torsello, A., and Cosmo, L. (2025). Generating graphs via spectral diffusion. In Intl. Conf. Learn. Repr. (ICLR)

    2025

  13. [21]

    G., and Segarra, S

    Navarro, M., Rey, S., Buciulea, A., Marques, A. G., and Segarra, S. (2024). Fair glasso: Estimating fair graphical models with unbiased statistical behavior. In Globerson, A., Mackey, L., Belgrave, D., Fan, A., Paquet, U., Tomczak, J., and Zhang, C., editors, Advances in Neura...

    2024

  14. [22]

    Niu, C., Song, Y ., Song, J., Zhao, S., Grover, A., and Ermon, S. (2020). Permutation invariant graph generation via score-based generative modeling. In Int. Conf. on Artif. Intell. and Stat. , pages 4474–4484. PMLR

    2020

  15. [23]

    Øksendal, B. (2003). Stochastic differential equations. Springer

    2003

  16. [24]

    M., Draxler, F., Karaletsos, T., and Mandt, S

    Pandey, K., Sofian, F. M., Draxler, F., Karaletsos, T., and Mandt, S. (2025). Variational control for guidance in diffusion models. arXiv preprint arXiv:2502.03686

    2025 arXiv

  17. [25]

    Pavon, M. (1989). Stochastic control and nonequilibrium thermodynamical systems. Applied Mathematics and Optimization, 19:187–202

    1989

  18. [26]

    Rout, L., Chen, Y ., Ruiz, N., Kumar, A., Caramanis, C., Shakkottai, S., and Chu, W.-S. (2025). Rb-modulation: Training-free personalization using stochastic optimal control. In Intl. Conf. Learn. Repr. (ICLR)

    2025

  19. [27]

    Rout, L., Raoof, N., Daras, G., Caramanis, C., Dimakis, A., and Shakkottai, S. (2024). Solving linear inverse problems provably via posterior sampling with latent diffusion models. Advances in Neural Inf. Process. Syst. (NeurIPS), 36

    2024

  20. [28]

    Sharma, K., Kumar, S., and Trivedi, R. (2024). Diffuse, sample, project: plug-and-play controllable graph generation. In Intl. Conf. on Machine Learning (ICML)

    2024

  21. [29]

    Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N., and Ganguli, S. (2015). Deep unsupervised learning using nonequilibrium thermodynamics. In Intl. Conf. on Machine Learning (ICML) , pages 2256–2265. PMLR

    2015

  22. [30]

    Song, J., Meng, C., and Ermon, S. (2020). Denoising diffusion implicit models. In Intl. Conf. Learn. Repr. (ICLR)

    2020

  23. [31]

    P., Kumar, A., Ermon, S., and Poole, B

    Song, Y ., Sohl-Dickstein, J., Kingma, D. P., Kumar, A., Ermon, S., and Poole, B. (2021). Score-based generative modeling through stochastic differential equations. In Intl. Conf. Learn. Repr. (ICLR). 11

    2021

  24. [32]

    Uehara, M., Zhao, Y ., Wang, C., Li, X., Regev, A., Levine, S., and Biancalani, T. (2025). Inference-time alignment in diffusion models with reward-guided generation: Tutorial and review. arXiv preprint arXiv:2501.09685

    2025 arXiv

  25. [33]

    Van Handel, R. (2007). Stochastic calculus, filtering, and stochastic control. Course notes., URL http://www. princeton. edu/rvan/acm217/ACM217. pdf, 14

    2007

  26. [34]

    Vignac, C., Krawczuk, I., Siraudin, A., Wang, B., Cevher, V ., and Frossard, P. (2023). Digress: Discrete denoising diffusion for graph generation. In Intl. Conf. Learn. Repr. (ICLR)

    2023

  27. [35]

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

    2011

  28. [36]

    Yang, N., Wu, H., Zeng, K., Li, Y ., Bao, S., and Yan, J. (2024). Molecule generation for drug design: a graph learning perspective. Fundamental Research

    2024

  29. [37]

    You, J., Liu, B., Ying, Z., Pande, V ., and Leskovec, J. (2018). Graph convolutional policy network for goal-directed molecular graph generation. Advances in Neural Inf. Process. Syst. (NeurIPS), 31

    2018

  30. [38]

    Zhou, C., Wang, X., and Zhang, M. (2024). Unifying generation and prediction on graphs with latent graph diffusion. Advances in Neural Inf. Process. Syst. (NeurIPS), 37:61963–61999

    2024

  31. [39]

    Zilberstein, N., Mardani, M., and Segarra, S. (2025). Repulsive latent score distillation for solving inverse problems. Intl. Conf. Learn. Repr. (ICLR)

    2025

  32. [40]

    reverse process

    Zilberstein, N., Sabharwal, A., and Segarra, S. (2024). Solving linear inverse problems using higher-order annealed Langevin diffusion. IEEE Trans. Signal Process. A HJB equation In this section, we give more details on our SOC formulation. The optimal control is given by U∗(G...

    2024

  33. [41]

    Ego-small: This dataset comprises 200 small ego graphs extracted from the larger Citeseer network

  34. [42]

    Community-small: Consisting of 100 synthetic graphs, this dataset features structures exhibiting distinct community partitions

  35. [43]

    Enzymes: We use the protein graphs from the BRENDA enzyme database, totaling 587 graphs. 14

  36. [44]

    These molecules are composed of up to 9 heavy atoms, including Carbon (C), Nitrogen (N), Oxygen (O), and Fluorine (F)

    QM9: A molecular dataset containing approximately 133,000 small molecules. These molecules are composed of up to 9 heavy atoms, including Carbon (C), Nitrogen (N), Oxygen (O), and Fluorine (F)

  37. [45]

    ZINC250k: This large molecular dataset includes 250,000 drug-like molecules. The graphs represent molecules with 6 to 38 heavy atoms, incorporating Carbon (C), Nitrogen (N), Oxygen (O), Fluorine (F), Phosphorus (P), Chlorine (Cl), Bromine (Br), and Iodine (I). E.2 Computationa...

  38. [46]

    to ensure a direct comparison of method performance under identical constraint settings, and are given by those values fulfilled by 10% of the graphs in the test dataset. The specific loss function used for each constraint is empirically selected from a pool of possibilities b...

  39. [47]

    The quantitative results for the fairness metrics and SBM validity for this scenario are presented in Table 3 in the main paper

    Random assignment: Sensitive attributes are assigned to nodes randomly. The quantitative results for the fairness metrics and SBM validity for this scenario are presented in Table 3 in the main paper. 15

  40. [48]

    This represents a more challenging scenario for generating fair graphs that are also valid Stochastic Block Models (SBMs)

    Community partitioning algorithm-based assignment: Sensitive attributes are assigned to nodes based on the community structure identified by a community partitioning algorithm. This represents a more challenging scenario for generating fair graphs that are also valid Stochasti...

    1931

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.