mlsanityData to 14 Oct 2026

Risk-Averse Inference-Time Diffusion Model Alignment with Finite-Sample Guarantees

arXiv 2610.17579 ↗First posted 13 Oct 2026stat.ML cs.LG math.OCcited by 0

On the map: Diffusion models → Diffusion foundations and sampling

also in 1 more
Abstract (arXiv)

Inference-time alignment steers diffusion models toward downstream objectives without costly fine-tuning, but existing methods typically optimize expected cost (or reward) without explicitly controlling tail risk. In this work, we propose a risk-averse framework for inference-time alignment of diffusion models that balances expected cost and Conditional Value-at-Risk (CVaR) under trajectory-level Kullback-Leibler regularization. We characterize the optimal path distribution and reverse transitions as functions of the CVaR threshold, reducing the alignment problem to a one-dimensional optimization problem over the threshold. We show that this one-dimensional optimization problem can be nonconvex in general, establish sufficient conditions for convexity, and introduce quadratic regularization to enable efficient derivative-free ternary search. Under suitable assumptions, we provide finite-sample guarantees for threshold optimization and value estimation, and combine them into a terminal total-variation bound relative to the regularized target under exact resampling. Experiments on synthetic data and image generation using Stable Diffusion demonstrate improved tail performance over risk-neutral baselines while maintaining competitive average cost.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 29 of its 36 references that are in the map. First posted on arXiv 13 Oct 2026.

Builds on

29 of 36 references.

Closest ancestors · 10

Same direction, newest first

  1. 2026Training-Free Adaptation of Diffusion Models via Doob's h-Transform18 Feb
  2. 2025Inference-Time Alignment in Diffusion Models with Reward-Guided Generation: Tutorial and Review16 Jan
  3. 2025A General Framework for Inference-time Scaling and Steering of Diffusion Models12 Jan
  4. 2024Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers17 Oct
  5. 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decoding15 Aug
  6. 2024Constrained Synthesis with Projected Diffusion Models5 Feb
  7. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  8. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  9. 2020Denoising Diffusion Implicit Models6 Oct
  10. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 19

  1. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
  2. 2022Diffusion Posterior Sampling for General Noisy Inverse Problemscited by 1,416
  3. 2023Training Diffusion Models with Reinforcement Learningcited by 982
  4. 2023DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Modelscited by 450
  5. 2023Directly Fine-Tuning Diffusion Models on Differentiable Rewardscited by 448
  6. 2023Practical and Asymptotically Exact Conditional Sampling in Diffusion Modelscited by 232
  7. 2024Improving Diffusion Inverse Problem Solving with Decoupled Noise Annealingcited by 163
  8. 2024Understanding Reinforcement Learning-Based Fine-Tuning of Diffusion Models: A Tutorial and Reviewcited by 98
  9. 2024Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Designcited by 79
  10. 2023SafeDiffuser: Safe Planning with Diffusion Probabilistic Modelscited by 75
  11. 2016Geometric Median in Nearly Linear Timecited by 39
  12. 2024Video Diffusion Models: A Surveycited by 30
7 more
  1. 2024Bridging Model-Based Optimization and Generative Modeling via Conservative Fine-Tuning of Diffusion Modelscited by 25
  2. 2025Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuningcited by 17
  3. 2025Constraint-Aware Diffusion Guidance for Robotics: Real-Time Obstacle Avoidance for Autonomous Racingcited by 11
  4. 2026Efficient Tail-Aware Generative Optimization via Flow Model Fine-Tuningcited by 8
  5. 2024Constrained Posterior Sampling: Time Series Generation with Hard Constraintscited by 7
  6. 2026On the Power of (Approximate) Reward Models for Inference-Time Scalingcited by 7
  7. 2026RS-Diffuser: Risk-Sensitive Diffusion Planning with Distributional Value Guidancecited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Inference-Time Alignment of Diffusion Models with Direct Noise Optimization
  2. 2025Inference-Time Alignment of Diffusion Models via Evolutionary Algorithms
  3. 2026Inference-Time Alignment for Diffusion Models via Variationally Stable Doob's Matching
  4. 2026Diffusion Alignment Beyond KL: Variance Minimisation as Effective Policy Optimiser
  5. 2026Inference-Time Attribute Distribution Alignment for Unconditional Diffusion
  6. 2025Direct Distributional Optimization for Provable Alignment of Diffusion Models
  7. 2025Test-time Alignment of Diffusion Models without Reward Over-optimization
  8. 2025Diffusion Alignment as Variational Expectation-Maximization