AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

ECHO: Entropy-Confidence Hybrid Optimization for Test-Time Reinforcement Learning

Chu Zhao, Enneng Yang, Yuting Liu, Jianzhe Zhao, Guibing Guo

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 EC-Tree entropy-confidence tree rollout (ablation)ECHO · AIME2025 (Qwen2.5-7B-Base, AIME2024 training)pass@16: 29.7 -> 54.3Improved
add EC-Tree entropy-confidence tree rollout (ablation)ECHO · AMC (Qwen2.5-7B-Base, AIME2024 training)pass@16: 54.2 -> 71.1Improved
add CA-Clip confidence-adaptive clipping (ablation)ECHO · AIME2025 (Qwen2.5-7B-Base, AIME2024 training)pass@16: 30.1 -> 54.3Improved
add E-SC Adv advantage shaping (ablation)ECHO · AIME2025 (Qwen2.5-7B-Base, AIME2024 training)pass@16: 53.5 -> 54.3Improved
add E-SC Adv advantage shaping (ablation)ECHO · AMC (Qwen2.5-7B-Base, AIME2024 training)pass@16: 78.7 -> 71.1Hurt
use ECHO instead of INTUITORtest-time RL · Qwen2.5-7B avg (AIME2024 training)pass@16 avg: 51.6 -> 55.3Improved
use ECHO instead of INTUITORtest-time RL · Qwen3-8B avg (AIME2024 training)pass@16 avg: 73.9 -> 76.1Improved
use ECHO instead of MM-UPTunsupervised multimodal RL · Qwen3-VL-8B avg (GeoQA training)pass@1 avg: 58.1 -> 60.5Improved
raise pruning threshold from 0.4 to 1.2 (ablation)ECHO · AIME2024Pass@16: 30.0 -> 16.7Hurt
lower pruning threshold from 0.4 to 0.2 (ablation)ECHO · AIME2024Pass@16: 30.0 -> 16.7Hurt
increase tail window from 2 to 8 (ablation)ECHO · AMCPass@16: 60.2 -> 75.9Improved
train on ReClor instead of math data (Qwen2.5-7B-Base)ECHO · MATH-500pass@16: 83.0 -> 81.9Hurt

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