AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

B-Spar: Bayesian Sparse-Reward Modeling for RL-based Image Editing

Shusong Xu, Peiye Liu, Yongbin Liu, Bangjie Yin, Tianyi Zheng, Zhaomang Sun, Zhenyu Chen, Peng-Tao Jiang, Jian Zhang, Yuzhao Wang, Zhen Gu, Jinwei Chen and 1 more

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning

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 Bayesian augmentation of positive samples before RL (ablation)B-Spar (RL only) · MIT-FiveK testHyperIQA: 64.83 -> 68.58Improved
add Gaussian reward densification on top of augmentation (ablation)B-Spar (RL only) · MIT-FiveK testLPIPS x100: 1.22 -> 1.17Improved
add SFT before RL (ablation)B-Spar · MIT-FiveK testHyperIQA: 68.63 -> 69.02Improved
use Bayesian reward modeling instead of GRPO with a learned reward model (ablation)B-Spar · MIT-FiveK testHyperIQA: 64.80 -> 69.02Improved
use Bayesian reward modeling instead of direct optimization (DPO) (ablation)B-Spar · MIT-FiveK testHyperIQA: 66.56 -> 69.02Improved
vary minimum improvement threshold delta in 1, 2, 3 (ablation)B-Spar · MIT-FiveK testHyperIQA: 68.75 / 69.02 / 68.69No clear effect
use overly wide Gaussian kernels (2.0) (ablation)B-Spar · MIT-FiveK testHyperIQA: slight drop in average performance, increased varianceHurt
use Bayesian-augmented rollout instead of vanilla offline rollout (ablation)B-Spar sampling · 5000 images action samplingreward proportion: 0.093 -> 0.236Improved
add dense reward signals in GRPO training (ablation)B-Spar · GRPO training reward curvemean reward: stagnant near zero -> approximately 0.3Improved
use B-Spar instead of other agents on fidelity to expert referenceB-Spar · MIT-FiveK testPSNR: 16.34 vs 20.75 (PhotoArt)Hurt

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