Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Adaptive Stochastic Coefficients for Accelerating Diffusion Sampling
Authors: Ruoyu Wang, Beier Zhu, Junzhi Li, Liangyu Yuan, Chi Zhang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate state-of-the-art performance: at 5 NFE, Ada SDE achieves FID scores of 4.18 on CIFAR-10, 8.05 on FFHQ and 6.96 on LSUN Bedroom. |
| Researcher Affiliation | Academia | 1AGI Lab, Westlake University 2Nanyang Technological University 3University of Chinese Academy of Sciences 4Institute of Software, Chinese Academy of Sciences 5 Tongji University |
| Pseudocode | Yes | Algorithm 1 Optimizing Θ1:N 1: Given: time schedules Tstu and Ttea 2: Repeat until convergence 3: Sample xt N = yt N N(0, t2 NI) 4: for n = N to 1 do 5: Sample ϵn N(0, I) 6: {γ, ξ, λ, µ}n Θn 7: ˆtn tn + γntn 8: xtn xtn + q ˆt2n t2nϵn 9: Compute xtn 1 using Eq. (9) 10: Update Θn via Eq. (10) 11: end for Algorithm 2 Ada SDE sampling 1: Given: parameters Θ1:N, student time schedule Tstu 2: Initialize xt N N(0, t2 NI) 3: for n = N to 1 do 4: Sample ϵn N(0, I) 5: {γ, ξ, λ, µ}n Θn 6: ˆtn tn + γntn 7: xtn xtn + q ˆt2n t2nϵn 8: Compute xtn 1 using Eq. (9) 9: end for 10: Return: xt0 |
| Open Source Code | Yes | Codes are available in https://github.com/Westlake-AGI-Lab/Ada SDE. |
| Open Datasets | Yes | We apply Ada SDE and DPM-Plugin to five pre-trained diffusion models across diverse domains. For pixel-space models, we include CIFAR10 (32 32) [27], FFHQ (64 64) [48], and Image Net (64 64) [49]. For latent-space models, we evaluate LSUN Bedroom (256 256) [50] with a guidance scale of 1.0. Additionally, we consider text-to-image high-resolution generation models, including Stable Diffusion v1.5 [5] at 512 512 resolution with a guidance scale of 7.5. |
| Dataset Splits | Yes | Sample quality is gauged using the Fréchet Inception Distance (FID) [54] over 50k images. For Stable-Diffusion, We evaluate FID as [54], using 30k samples from fixed prompts based on the MS-COCO [28] validation set. The random seed was fixed to 0 to ensure consistent reproducibility of the experimental results. |
| Hardware Specification | Yes | We conducted our experiments using NVIDIA A800 and 4090 GPUs with 80GB and 24GB memory, respectively. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, CUDA versions). |
| Experiment Setup | Yes | Our Ada SDE is assessed at low NFE settings (NFE {3, 5, 7, 9}) with AFS [53] implemented. Sample quality is gauged using the Fréchet Inception Distance (FID) [54] over 50k images. For Stable-Diffusion, We evaluate FID as [54], using 30k samples from fixed prompts based on the MS-COCO [28] validation set. The random seed was fixed to 0 to ensure consistent reproducibility of the experimental results. For tuning across all datasets, we employed a learning rate of 0.2 along with a cosine learning rate schedule (coslr). |