ICML 2026

Triadic Dynamics Aware Diffusion Posterior
Sampling for Inverse Problems:

Optimizing Guidance and Stochasticity Schedules

Junseo Bang1,*   Dong Ju Mun1,*   Hoigi Seo1   Seongmin Hong2   Se Young Chun1,2,3

1Dept. of Electrical & Computer Engineering  路  2INMC  路  3IPAI & AIIS, Seoul National University, Republic of Korea

*Equal contribution  路  Correspondence: sychun@snu.ac.kr

arXiv Code BibTeX 馃HF Paper

Key Idea

Posterior sampling as an optimization problem of
time-varying schedules over three coupled components

Generative posterior sampling for inverse problems usually consists of three main components: data consistency (DC) guidance, classifier-free guidance (CFG), and stochasticity. While prior works have focused on how to develop each or all of these components, far less attention has been paid to how to schedule them, leading to heuristically fixed or only partially adjusted, suboptimal schedules. We reveal that these three components follow triadic coupling dynamics, and propose TriPS (Triadic Dynamics Aware Diffusion Posterior Sampling), which optimizes the time-varying schedules of these three coupled forces.

TriPS key idea: the three key components and the triadic coupling (DC-CFG conflict)

Qualitative Samples

Linear inverse problems (SD3.5-M)

1–3 / 15

Drag the handle on any pair to reveal each side. Inpainting uses TriPS-T; the others use TriPS-G.

Analysis

Triadic Coupling Dynamics

We analyze the triadic coupling dynamics in posterior sampling, governed by DC guidance, CFG, and stochasticity. We formalize the early-stage DC-CFG conflict and show how stochasticity regularizes sampling trajectories toward higher-probability regions.

Early stage guidance conflict
Early-stage guidance conflict. Since DC guidance and CFG originate from distinct objectives, their update directions are misaligned. As the CFG scale increases, this early-stage conflict intensifies, slowing the decay of the measurement residual and inducing semantic hallucinations that deviate from the measurement.
Stochasticity as a regularizer
Stochasticity as a regularizer. Stronger DC guidance or CFG drives sampling away from higher-probability trajectories, lowering its alignment with the score function, whereas appropriately scaled stochasticity consistently improves this alignment, suppressing artifacts and restoring perceptual fidelity.

➤ Triadic Scheduling Trend

β(t)↓DC guidance
λ(t)↑Classifier-free guidance
η(t)↓Stochasticity

Method

Triadic Schedule Optimization

To realize the triadic scheduling trend, TriPS comprises two complementary paradigms: a template-based schedule search that identifies robust schedule curves from a discrete family of functional forms, and a GRPO-based schedule optimization that captures complex temporal curves beyond the fixed functional templates.

TriPS method overview
(Left) TriPST explores a discrete search space defined by compact templates (Linear, Exp, Log) that satisfy the triadic scheduling trend (尾(t)↓, 位(t)↑, 畏(t)↓), turning a high-dimensional schedule search into a low-dimensional selection. (Right) TriPSG enables continuous schedule discovery beyond fixed functional forms: a policy 蟺胃 samples coefficients from learnable Beta distributions to parameterize the schedules via Bernstein polynomials, strictly constraining each curve within a valid range, and is updated by rewards derived from the restored images.

Quantitative Comparison

Linear inverse problems (SD3.5-M)

Table 1: quantitative comparison
Quantitative comparison on linear inverse problems with the SD3.5-M flow matching prior (Table 1 of the paper).

Ablation

Reward-guided Perception–Distortion Control

Reward guided perception distortion control
By reweighting the distortion and perception terms of the hybrid reward, TriPSG navigates the perception-distortion trade-off. The optimized schedules align with the triadic scheduling trend while exhibiting fine-grained temporal variations, such as local non-monotonic fluctuations and magnitude shifts, that prove critical for pushing the perception-distortion Pareto frontier.

Cite

BibTeX

@misc{bang2026triadicdynamicsawarediffusion,
  title         = {Triadic Dynamics Aware Diffusion Posterior Sampling for Inverse
                   Problems: Optimizing Guidance and Stochasticity Schedules},
  author        = {Junseo Bang and Dong Ju Mun and Hoigi Seo and Seongmin Hong and Se Young Chun},
  year          = {2026},
  eprint        = {2605.26470},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV},
  url           = {https://arxiv.org/abs/2605.26470}
}