Diffusion Alignment Beyond KL: Variance Minimisation as Effective Policy Optimiser

• URL: http://arxiv.org/abs/2602.12229v1
• Date: Thu, 12 Feb 2026 18:06:03 GMT
• Title: Diffusion Alignment Beyond KL: Variance Minimisation as Effective Policy Optimiser
• Authors: Zijing Ou, Jacob Si, Junyi Zhu, Ondrej Bohdal, Mete Ozay, Taha Ceritli, Yingzhen Li,
• Abstract summary: We introduce Variance Minimisation Policy optimisation (VMPO)<n>VMPO formulates diffusion alignment as minimising the variance of log importance weights.<n>Under different choices of potential functions and variance minimisation strategies, VMPO recovers various existing methods.
• Score: 35.260598916253635
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Diffusion alignment adapts pretrained diffusion models to sample from reward-tilted distributions along the denoising trajectory. This process naturally admits a Sequential Monte Carlo (SMC) interpretation, where the denoising model acts as a proposal and reward guidance induces importance weights. Motivated by this view, we introduce Variance Minimisation Policy Optimisation (VMPO), which formulates diffusion alignment as minimising the variance of log importance weights rather than directly optimising a Kullback-Leibler (KL) based objective. We prove that the variance objective is minimised by the reward-tilted target distribution and that, under on-policy sampling, its gradient coincides with that of standard KL-based alignment. This perspective offers a common lens for understanding diffusion alignment. Under different choices of potential functions and variance minimisation strategies, VMPO recovers various existing methods, while also suggesting new design directions beyond KL.

Related papers

• Variational Optimality of Föllmer Processes in Generative Diffusions [20.583125441867434]
  We analyze generative diffusions that transport a point mass to a prescribed target distribution over a finite time horizon.<n>We show that the diffusion coefficient can be tuned emphaposteriori without changing the time-marginal distributions.
  arXiv  Detail & Related papers  (2026-02-11T16:15:19Z)
• SetPO: Set-Level Policy Optimization for Diversity-Preserving LLM Reasoning [50.93295951454092]
  We introduce a set level diversity objective defined over sampled trajectories using kernelized similarity.<n>Our approach derives a leave-one-out marginal contribution for each sampled trajectory and integrates this objective as a plug-in advantage shaping term for policy optimization.<n>Experiments across a range of model scales demonstrate the effectiveness of our proposed algorithm, consistently outperforming strong baselines in both Pass@1 and Pass@K across various benchmarks.
  arXiv  Detail & Related papers  (2026-02-01T07:13:20Z)
• An Elementary Approach to Scheduling in Generative Diffusion Models [55.171367482496755]
  An elementary approach to characterizing the impact of noise scheduling and time discretization in generative diffusion models is developed.<n> Experiments across different datasets and pretrained models demonstrate that the time discretization strategy selected by our approach consistently outperforms baseline and search-based strategies.
  arXiv  Detail & Related papers  (2026-01-20T05:06:26Z)
• A Diffusion Model Framework for Maximum Entropy Reinforcement Learning [32.26181994745642]
  We present a modified surrogate objective for MaxEntRL that incorporates diffusion dynamics in a principled way.<n>We find that DiffSAC, DiffPPO and DiffWPO achieve better returns and higher sample efficiency than SAC and PPO.
  arXiv  Detail & Related papers  (2025-12-01T18:59:58Z)
• Training-Free Stein Diffusion Guidance: Posterior Correction for Sampling Beyond High-Density Regions [46.59494117137471]
  Training free diffusion guidance provides a flexible way to leverage off-the-shelf classifiers without additional training.<n>We introduce Stein Diffusion Guidance (SDG), a novel training-free framework grounded in a surrogate SOC objective.<n>Experiments on molecular low-density sampling tasks suggest that SDG consistently surpasses standard training-free guidance methods.
  arXiv  Detail & Related papers  (2025-07-07T21:14:27Z)
• Aligning Latent Spaces with Flow Priors [72.24305287508474]
  This paper presents a novel framework for aligning learnable latent spaces to arbitrary target distributions by leveraging flow-based generative models as priors.<n> Notably, the proposed method eliminates computationally expensive likelihood evaluations and avoids ODE solving during optimization.
  arXiv  Detail & Related papers  (2025-06-05T16:59:53Z)
• Importance Weighted Score Matching for Diffusion Samplers with Enhanced Mode Coverage [16.94974733994214]
  prevailing methods often circumvent the lack of target data by optimizing reverse KL-based objectives.<n>We propose a principled approach for training diffusion-based samplers by directly targeting an objective analogous to the forward KL divergence.<n>Our approach consistently outperforms existing neural samplers across all distributional distance metrics.
  arXiv  Detail & Related papers  (2025-05-26T02:48:26Z)
• Aligning Few-Step Diffusion Models with Dense Reward Difference Learning [81.85515625591884]
  Stepwise Diffusion Policy Optimization (SDPO) is an alignment method tailored for few-step diffusion models. SDPO incorporates dense reward feedback at every intermediate step to ensure consistent alignment across all denoising steps. SDPO consistently outperforms prior methods in reward-based alignment across diverse step configurations.
  arXiv  Detail & Related papers  (2024-11-18T16:57:41Z)
• Amortized Posterior Sampling with Diffusion Prior Distillation [55.03585818289934]
  Amortized Posterior Sampling is a novel variational inference approach for efficient posterior sampling in inverse problems.<n>Our method trains a conditional flow model to minimize the divergence between the variational distribution and the posterior distribution implicitly defined by the diffusion model.<n>Unlike existing methods, our approach is unsupervised, requires no paired training data, and is applicable to both Euclidean and non-Euclidean domains.
  arXiv  Detail & Related papers  (2024-07-25T09:53:12Z)
• Sequential Monte Carlo for Inclusive KL Minimization in Amortized Variational Inference [3.126959812401426]
  We propose SMC-Wake, a procedure for fitting an amortized variational approximation that uses sequential Monte Carlo samplers to estimate the gradient of the inclusive KL divergence. In experiments with both simulated and real datasets, SMC-Wake fits variational distributions that approximate the posterior more accurately than existing methods.
  arXiv  Detail & Related papers  (2024-03-15T18:13:48Z)
• Adaptive Annealed Importance Sampling with Constant Rate Progress [68.8204255655161]
  Annealed Importance Sampling (AIS) synthesizes weighted samples from an intractable distribution. We propose the Constant Rate AIS algorithm and its efficient implementation for $alpha$-divergences.
  arXiv  Detail & Related papers  (2023-06-27T08:15:28Z)
• Variational Refinement for Importance Sampling Using the Forward Kullback-Leibler Divergence [77.06203118175335]
  Variational Inference (VI) is a popular alternative to exact sampling in Bayesian inference. Importance sampling (IS) is often used to fine-tune and de-bias the estimates of approximate Bayesian inference procedures. We propose a novel combination of optimization and sampling techniques for approximate Bayesian inference.
  arXiv  Detail & Related papers  (2021-06-30T11:00:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.