Pith. sign in

REVIEW 18 cited by

Feynman-Kac Correctors in Diffusion: Annealing, Guidance, and Product of Experts

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2503.02819 v2 pith:H7FKSFSA submitted 2025-03-04 cs.LG

Feynman-Kac Correctors in Diffusion: Annealing, Guidance, and Product of Experts

classification cs.LG
keywords modelsdistributionsfeynman-kacguidanceinference-timemethodspretrainedsampling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

While score-based generative models are the model of choice across diverse domains, there are limited tools available for controlling inference-time behavior in a principled manner, e.g. for composing multiple pretrained models. Existing classifier-free guidance methods use a simple heuristic to mix conditional and unconditional scores to approximately sample from conditional distributions. However, such methods do not approximate the intermediate distributions, necessitating additional `corrector' steps. In this work, we provide an efficient and principled method for sampling from a sequence of annealed, geometric-averaged, or product distributions derived from pretrained score-based models. We derive a weighted simulation scheme which we call Feynman-Kac Correctors (FKCs) based on the celebrated Feynman-Kac formula by carefully accounting for terms in the appropriate partial differential equations (PDEs). To simulate these PDEs, we propose Sequential Monte Carlo (SMC) resampling algorithms that leverage inference-time scaling to improve sampling quality. We empirically demonstrate the utility of our methods by proposing amortized sampling via inference-time temperature annealing, improving multi-objective molecule generation using pretrained models, and improving classifier-free guidance for text-to-image generation. Our code is available at https://github.com/martaskrt/fkc-diffusion.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 18 Pith papers

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

  1. A Priori Sampling of Transition States with Guided Diffusion

    physics.chem-ph 2026-03 conditional novelty 8.0

    ASTRA reframes transition-state search as guided diffusion inference that samples the isodensity surface between metastable basins and converges to first-order saddles via score differences and physical forces.

  2. Bootstrap Flow-Map Tree Sampling Enables Online Feedback Driven Search

    cs.LG 2026-07 conditional novelty 7.0

    Bootstrap Flow-Map Trees construct complete DDPM-like trajectories with a single NFE and dynamic steps, enabling efficient online feedback-driven search and alignment that beats prior tree and SMC samplers.

  3. Adaptive Order Policies for Masked Diffusion

    cs.LG 2026-05 unverdicted novelty 7.0

    A policy network learns to choose unmasking order in masked diffusion by reweighting the loss, outperforming random and heuristic baselines on ordering-sensitive tasks.

  4. Contrastive Distribution Matching for Amortized Sequential Monte Carlo in Discrete Diffusion

    cs.LG 2026-05 unverdicted novelty 7.0

    CDM amortizes SMC inference for reward-tilted discrete diffusion by training a parameterized twist function on contrastive samples with closed-form kernels.

  5. SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate

    stat.ML 2026-05 unverdicted novelty 7.0

    URGE performs unbiased path-wise importance reweighting via Girsanov estimation for derivative-free inference-time scaling in diffusion models, proving equivalence to particle-wise SMC and outperforming baselines empirically.

  6. Diversify Diffusion with Temperature Sampling and Variance-Corrective Time Shifting

    cs.CV 2026-07 conditional novelty 6.0

    Variance-corrective time shifting cancels the variance inflation of naive temperature sampling so pretrained diffusion models can raise diversity without retraining.

  7. Sequentially-Controlled Interactive Multi-Particle Flow-Maps for Online Feedback-Driven Search

    cs.LG 2026-07 unverdicted novelty 6.0

    IMPFM is a multi-particle flow-map sampling method with sequential posterior sharing and interaction-aware correction that targets a KL-tilted distribution for global exploration in online feedback search.

  8. Parallel Tempering Initial Sampling in Inference-Time Reward Alignment

    cs.LG 2026-05 unverdicted novelty 6.0

    PATHS applies parallel tempering to improve initial particle sampling for SMC reward alignment, yielding better results on layout-to-image and quantity-aware generation tasks.

  9. SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate

    stat.ML 2026-05 unverdicted novelty 6.0

    SURGE is an unbiased particle filter that fuses diffusion-model simulations with noisy observations via sequential Monte Carlo reweighting over diffusion trajectories.

  10. Simple Approximation and Derivative Free Inference-Time Scaling for Diffusion Models via Sequential Monte Carlo on Path Measures

    stat.ML 2026-05 unverdicted novelty 6.0

    URGE performs unbiased inference-time scaling for diffusion models by attaching multiplicative path weights from Girsanov estimation and resampling trajectories, with a proven equivalence to prior particle-wise SMC schemes.

  11. Optimizing Diffusion Priors in Image Reconstruction from a Single Observation

    cs.CV 2026-04 unverdicted novelty 6.0

    Combining diffusion priors as a product-of-experts and optimizing exponents via Bayesian evidence maximization enables prior tuning from one observation in inverse imaging problems.

  12. Generative optimal transport via forward-backward HJB matching

    cond-mat.stat-mech 2026-04 unverdicted novelty 6.0

    A forward-backward HJB duality computes the optimal stochastic transport control from easy forward relaxation trajectories alone, expressed as path-space free energy without backward simulation.

  13. Momentum Guidance: Plug-and-Play Guidance for Flow Models

    cs.LG 2026-02 conditional novelty 6.0

    Momentum Guidance improves flow-model sample quality by extrapolating the current velocity away from an exponential moving average of past velocities, with no extra model evaluations.

  14. Enhanced Diffusion Sampling: Efficient Rare Event Sampling and Free Energy Calculation with Diffusion Models

    stat.ML 2026-02 conditional novelty 6.0

    Steering a pretrained diffusion model with bias potentials and then reweighting with MBAR computes rare-state free energies with far fewer samples than unbiased diffusion sampling.

  15. CMAD: Cooperative Multi-Agent Diffusion via Stochastic Optimal Control

    cs.LG 2026-02 unverdicted novelty 6.0

    CMAD formulates compositional generation as cooperative stochastic optimal control among pre-trained diffusion models, validated on conditional MNIST against a gradient-guidance baseline.

  16. On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization

    cs.LG 2026-07 conditional novelty 5.5

    Online fine-tuning of discrete diffusion models with complementary acquisition, CVaR shaping, density-entropy debiasing, replay, and validity control finds better molecules under fixed oracle budgets than offline fine...

  17. Local MAP Sampling for Diffusion Models

    cs.GR 2025-10 conditional novelty 4.0

    LMAPS frames reverse-diffusion inverse-problem solving as repeated local MAP estimation, unifying existing optimization-based solvers, and achieves strong PSNR gains on tasks like motion deblurring, JPEG restoration, ...

  18. Inference-Time Alignment Control for Diffusion Models with Reinforcement Learning Guidance

    cs.LG 2025-08 conditional novelty 4.0

    Blending a base diffusion model with its RL-finetuned version at sampling time lets users dial alignment strength, with the blend weight corresponding to the KL-regularization coefficient beta/w.