AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

EAPO: Enhancing Policy Optimization with On-Demand Expert Assistance

Siyao Song, Cong Ma, Zhihao Cheng, Shiye Lei, Minghao Li, Ying Zeng, Huaixiao Tou, Kai Jia

ICML 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants · 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 on-demand expert assistance in RL (EAPO) vs self-exploratory RLR1-Distill-7B RL · Math averagePass@32: 59.16 -> 64.07Improved
add expert assistance in RL on Qwen3-8BQwen3-8B RL · Math averagePass@32: 45.44 -> 49.98Improved
use parallel instead of sequential expert queries (ablation)EAPO · Math averagePass@32: 61.79 -> 64.07Improved
use 32B expert instead of 14B expert (ablation)EAPO · Math averagePass@32: 61.55 -> 64.07Improved
use heterogeneous Llama-8B experts instead of homogeneous R1-distill-7B experts (ablation)EAPO · Math averagepass@32: 58.85 -> 60.66Improved
add homogeneous expert pool to self-exploratory RL (ablation)EAPO · Math averagepass@32: 59.16 -> 58.85No clear effect
use annealed schedule s^-1 instead of budget-matched fixed rate 0.02 (ablation)EAPO · Math averagePass@32: 59.65 -> 64.07Improved
use faster decay s^-2 instead of s^-1 (ablation)EAPO · Math averagePass@32: 64.07 -> 59.76Hurt
use informative random-sampled expert guidance instead of error-sampled guidance (ablation)Expert-Assisted Workflow · Math averagepass@32: 42.82 -> 49.39Improved
use real expert guidance instead of meaningless guidance (ablation)Expert-Assisted Workflow · Math averagepass@32: 42.92 -> 49.39Improved
self collaboration with 7B copiesR1-7B · Math averageaccuracy: 42.53 -> 41.83No clear effect
use stronger expert (DeepSeek_V3.2) over Qwen3-235B-A22B in workflow (ablation)Expert-Assisted Workflow · Math averageaccuracy: 52.61 -> 54.55Improved

Later papers that compared with this method

1 later accepted paper reports EAPO as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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