AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

RLAnything: Forge Environment, Policy, and Reward Model in Completely Dynamic RL System

Yinjie Wang, Tianbao Xie, Ke Shen, Mengdi Wang, Ling Yang

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
add dynamic reward model optimization (consistency feedback) on top of policy training (ablation)RLAnything GUI agent · OSWorld in-domainaccuracy: 35.3 -> 36.9Improved
add environment adaptation on top of policy + reward (ablation)RLAnything GUI agent · OSWorld in-domainaccuracy: 36.9 -> 40.6Improved
add environment adaptation on top of policy + reward (ablation)RLAnything LLM agent · AlfWorld in-domainaccuracy: 55.6 -> 60.2Improved
add environment adaptation to policy + reward (ablation)reward model · AlfWorld process accuracyprocess accuracy: 55.2 -> 59.4Improved
add environment adaptation to policy + reward (ablation)coding unit-test generator · LiveBench Detectdetect accuracy: 37.9 -> 48.5Improved
add joint reward model training to policy (ablation)coding LLM · LiveBench UT accuracyUT accuracy: 27.8 -> 73.3Improved
add environment adaptation (ablation)reward model outcome accuracy in GUI/LLM agents · AlfWorld outcome accuracyoutcome accuracy: 61.1 -> 61.7Improved
use integrated step-wise plus outcome reward instead of outcome-only (GRPO) (ablation)policy training · AlfWorld and OSWorld training curvestraining reward curveImproved
train policy with only optimized reward model signal, no evaluator-script outcome (ablation)GUI policy · OSWorld training curvetraining curveImproved
vary lambda from 1 to 1/4 in policy reward (ablation)RLAnything policy · AlfWorldpolicy accuracy: 54.1 -> 48.0Hurt
increase lambda from 1 to 4 (ablation)RLAnything policy · AlfWorldpolicy accuracy: 54.1 -> 53.3Hurt
add self-consistency signal (lambda=1 vs 0) to reward model training (ablation)reward model · AlfWorldstep accuracy: 48.2 -> 54.7Improved

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