AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

T$^2$PO: Uncertainty-Guided Exploration Control for Stable Multi-Turn Agentic Reinforcement Learning

Haixin Wang, Hejie Cui, Chenwei Zhang, Xin Liu, Shuowei Jin, Shijie Geng, Xinyang Zhang, Nasser Zalmout, Zhenyu Shi, Yizhou Sun

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 rejective fine-tuning cold start (ablation)T2PO · WebShopsuccess rate: 61.32 -> 81.64Improved
add token-level thinking intervention (ablation)T2PO · WebShopsuccess rate: 73.27 -> 81.64Improved
add turn-level dynamical sampling (ablation)T2PO · WebShoptask score: 72.40 -> 93.84Improved
add turn-level dynamical sampling (ablation)T2PO · WebShopsuccess rate: 63.67 -> 81.64Improved
use TTI+TDS instead of a lengthy reward (ablation)T2PO · WebShopsuccess rate: 65.87 -> 81.64Improved
use TTI+TDS instead of short CoT cold-start (ablation)T2PO · WebShopsuccess rate: 71.29 -> 81.64Improved
use TTI+TDS instead of void turn filtering (ablation)T2PO · WebShopsuccess rate: 76.20 -> 81.64Improved
use TTI+TDS instead of a hard thinking budget (ablation)T2PO · WebShopsuccess rate: 79.21 -> 81.64Improved
use T2PO instead of GiGPO on Qwen3-4B-RFTGiGPO · ALFWorldsuccess rate: 80.47 -> 90.23Improved
use T2PO instead of GiGPO on Qwen3-8B-RFTGiGPO · WebShopsuccess rate: 77.92 -> 82.42Improved
use T2PO instead of GiGPO on search QA (Qwen3-4B)GiGPO · Search QA averageaverage score: 52.97 -> 54.93Improved
increase rollout batch size from 8 to 16 (group size 8) (ablation)T2PO · WebShopsuccess rate: 80.97 -> 81.64No 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