AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Inpainting-Guided Policy Optimization for Diffusion Large Language Models

Siyan Zhao, Mengchen Liu, Jing Huang, Miao Liu, Chenyu Wang, Bo Liu, Yuandong Tian, Guan Pang, Sean Bell, Aditya Grover, Feiyu Chen

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 26 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 IGPO RL on LLaDA-Instruct (vs DiffuGRPO)LLaDA-8B-Instruct · GSM8Kpass@1: 81.9 -> 83.6Improved
add IGPO RL on LLaDA-Instruct (vs DiffuGRPO)LLaDA-8B-Instruct · MATH500pass@1: 40.2 -> 42.8Improved
add IGPO RL on LLaDA-Instruct (vs DiffuGRPO)LLaDA-8B-Instruct · AMCavg@16: 17.5 -> 18.1Improved
add Length-aligned SFT on rewritten traces (ablation)LLaDA-8B-Instruct · MATH500pass@1: 39.0 -> 45.2Improved
add Length-aligned SFT on rewritten traces (ablation)LLaDA-8B-Instruct · GSM8Kpass@1: 81.5 -> 83.6Improved
add IGPO after Length-aligned SFT (ablation)LLaDA-8B-Instruct + Length-aligned SFT · GSM8Kpass@1: 83.6 -> 86.4Improved
add IGPO after Length-aligned SFT (ablation)LLaDA-8B-Instruct + Length-aligned SFT · MATH500pass@1: 45.2 -> 47.4Improved
add IGPO after Length-aligned SFT (ablation)LLaDA-8B-Instruct + Length-aligned SFT · AMCavg@16: 22.3 -> 24.4Improved
add inpainting hints to GRPO sampling (ablation)GRPO on LLaDA · training all-wrong group ratiofraction of all-wrong groups: reduced by around 60%Improved
use partial hint injection (eta in U[0.2,0.6]) instead of full ground-truth injection (eta=1.0) (ablation)IGPO · GSM8K, MATH500, AMCaccuracyImproved
add hint injection versus no hints (ablation)GRPO on LLaDA · GSM8K, MATH500, AMCaccuracyImproved
restrict hint-token gradients to top 20% entropy positions (tau=0.2 vs 1.0) (ablation)IGPO · GSM8K (temp 0.1, length 256)accuracyImproved

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