AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Learning to Explore: Scaling Agentic Reasoning via Exploration-Aware Policy Optimization

Xingyuan Hua, Sheng Yue, Ju Ren

ICML 2026 · Reinforcement learning › RL from feedback · approach: Scaling RL with verifiable rewards · 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
naive test-time exploration added to GRPOGRPO Qwen-VL-2B · AndroidWorldsuccess rate: 52.3 -> 55.1Improved
naive test-time exploration added to GRPOGRPO Qwen-VL-8B · AndroidWorldsuccess rate: 55.4 -> 52.9Hurt
naive test-time exploration added to DAPODAPO Qwen-VL-2B · AndroidWorldsuccess rate: 56.8 -> 54.6Hurt
naive test-time exploration added to GiGPOGiGPO Qwen-VL-2B · AndroidWorldsuccess rate: 56.0 -> 53.1Hurt
naive test-time exploration added to LAMERLAMER Qwen-VL-2B · AndroidWorldsuccess rate: 61.2 -> 56.3Hurt
naive test-time exploration added to LAMERLAMER Qwen-VL-4B · AndroidWorldsuccess rate: 61.5 -> 58.2Hurt
naive test-time exploration added to LAMERLAMER Qwen-VL-8B · AndroidWorldsuccess rate: 61.0 -> 57.3Hurt
add exploratory reward (ablation)Ours · all four environmentssuccess rateImproved
add exploration-aware grouping (ablation)Ours · all four environmentssuccess rateImproved
add format reward (ablation)Ours · all four environmentssuccess rateImproved
train on AndroidWorld and test on unseen OSWorldOurs 8B · OSWorld overallsuccess rate: 60.64Improved
increase discount factor gamma toward 1.0 (ablation)Ours · text and GUI environmentssuccess rateMixed

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