AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Diffusion Controller: Framework, Algorithms and Parameterization

Tong Yang, Moonkyung Ryu, Chih-Wei Hsu, Guy Tennenholtz, Yuejie Chi, Craig Boutilier, Bo Dai

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models · 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
use LS-MDP-motivated split parameterization instead of naive side network (ablation)DiffCon (gray-box) · SFT, SD v1.4HPS-v2 win rate: 0.5655 -> 0.6667Improved
use LS-MDP-motivated split parameterization instead of naive side network (ablation)DiffCon (gray-box) · RWL, SD v1.4HPS-v2 win rate: 0.5060 -> 0.6815Improved
use LS-MDP-motivated split parameterization instead of naive side network (ablation)DiffCon (gray-box) · PPO, SD v1.4HPS-v2 win rate: 0.5201 -> 0.6957Improved
replace white-box LoRA with gray-box DiffCon (fewer parameters)DiffCon · SFT, SD v1.4HPS-v2 win rate: 0.5766 -> 0.6667Improved
replace white-box LoRA with gray-box DiffConDiffCon · RWL, SD v1.4HPS-v2 win rate: 0.6109 -> 0.6815Improved
replace white-box LoRA with gray-box DiffConDiffCon · PPO, SD v1.4HPS-v2 win rate: 0.9048 -> 0.6957Hurt
add DiffCon on top of LoRA (separate training, DiffCon-S)LoRA · PPO, SD v1.4HPS-v2 win rate: 0.9048 -> 0.9315Improved
add DiffCon on top of LoRA (joint training, DiffCon-J)LoRA · RWL, SD v1.4HPS-v2 win rate: 0.6109 -> 0.6555Improved
add DiffCon on top of LoRA (separate training, DiffCon-S)LoRA · RWL, SD v1.4HPS-v2 win rate: 0.6109 -> 0.7091Improved
increase test-time side-network guidance strength (ablation)DiffCon · SD v1.4 text-to-imagegeneration qualityImproved

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