AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

EchoRL: Reinforcement Learning via Rollout Echoing

Jinhe Bi, Aniri, Minglai Yang, Xingcheng Zhou, Wenke Huang, Sikuan Yan, Yujun Wang, Zixuan Cao, Michael Färber, Xun Xiao, Volker Tresp, Yunpu Ma

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 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 EchoRL auxiliary loss to GRPOGRPO, Qwen2.5-Math-7B · math ID averageavg accuracy: 44.5 -> 47.0Improved
add EchoRL to DAPODAPO, Qwen2.5-Math-7B · OOD averageavg accuracy: 57.6 -> 60.9Improved
add EchoRL to LUFFYLUFFY, Qwen2.5-Math-7B · math ID averageavg accuracy: 50.1 -> 51.9Improved
add EchoRL to UFTUFT, Qwen2.5-Math-7B · math ID averageavg accuracy: 45.7 -> 47.6Improved
add EchoRL to GRPOGRPO, Qwen2.5-Math-7B · AIME24avg@32: 25.8 -> 24.9Hurt
add EchoRL to LUFFYLUFFY, Qwen2.5-Math-7B · Olympiadaccuracy: 57.2 -> 55.1Hurt
add EchoRL to GRPOGRPO · training entropy, Qwen2.5-Math-7Bstep-level entropyImproved
set auxiliary loss weight lambda to 0.1 or larger instead of 0.001 (ablation)EchoRL · Qwen2.5-Math-7BaccuracyHurt
supervise EchoClip (prefix to max-entropy step) instead of full rollout or low-entropy clip (ablation)EchoRL · Appendix C ablationaccuracyImproved
activate EchoRL late in training instead of full training (ablation)EchoRL · DAPO vs GRPOaccuracyMixed
remove reasoning steps by highest entropy first vs lowest-first or random (ablation)OpenR1-Math 45k trajectories · step removal analysisaccuracyHurt
add EchoRL to GRPOGRPO · mean response lengthresponse lengthNo clear effect

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