AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

Fisher-Preserving Guidance: Training-Free Manifold Constraints for Safe Diffusion Control

Hao Ren, Zetong Bi, Yiming Zeng, Le Zheng, Zhi Li, Zhaoliang Wan, Lu Qi, Hui Cheng

ICML 2026 · Image generation › Fast diffusion sampling · approach: Guidance for conditional diffusion · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Fisher-preserving guidance to NoMaDNoMaD · CARLAsuccess rate (%): 51.11 -> 60.00Improved
add uncertainty-guided action blending to NoMaDNoMaD · CARLAsuccess rate (%): 51.11 -> 57.78Improved
add both FPG and action blending (ablation)NoMaD · CARLAsuccess rate (%): 48.89 -> 75.55Improved
add FPG on top of action blending (ablation)NoMaD with AB · CARLAsuccess rate (%): 57.78 -> 75.55Improved
add action blending on top of FPG (ablation)NoMaD with FPG · CARLAaverage collisions: 0.911 -> 0.445Improved
use uncertainty-guided blending instead of random blending (ablation)FPG + AB · CARLAsuccess rate (%): 57.78 -> 75.55Improved
add FPG to the Maze2D sampler without task guidanceDDPM baseline · Maze2Dcollisions: 0.243 -> 0.170Improved
add FPG on top of TSDF guidanceDDPM baseline + TG · Maze2Dcollisions: 0.071 -> 0.016Improved
add FPG to official Diffusion PolicyDiffusion Policy · PushTscore: 0.91 -> 0.94Improved
add FPG on real robotNoMaD · Real robot, 3 scenessuccess rate (successes/30): 19/30 -> 28/30Improved
use OPS low-rank projection instead of VJP-Fisher (ablation)Fisher guidance · RTX 3060, 10 stepstotal latency (ms): 95.32 -> 61.93Improved
reduce denoising steps from 10 to 4 with OPS guidance (ablation)FPG-OPS · RTX 3060total latency (ms): 61.93 -> 45.12Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy