AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Information Gain-based Policy Optimization: A Simple and Effective Approach for Multi-Turn LLM Agents

Guoqing Wang, Sunhao Dai, Guangze Ye, Zeyu Gan, Wei Yao, Yong Deng, Xiaofeng Wu, Zhenzhe Ying

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning

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 turn-level information gain reward to outcome reward (ablation)GRPO on Qwen2.5-7B-Instruct · 7 QA datasets avgF1: 51.9 -> 60.2Improved
add turn-level information gain reward to outcome reward (ablation)GRPO on Qwen2.5-3B-Instruct · 7 QA datasets avgF1: 32.3 -> 48.9Improved
use information gain reward only, no outcome reward (ablation)IGPO on Qwen2.5-7B-Instruct · 7 QA datasets avgF1: 53.5 vs 51.9 (GRPO), 60.2 (full)Mixed
use information gain only vs outcome F1 only (ablation)IGPO on Qwen2.5-3B-Instruct · 7 QA datasets avgF1: 32.3 -> 34.6Improved
separate normalization of IG and outcome rewards vs joint (ablation)IGPO on Qwen2.5-7B-Instruct · 7 QA datasets avgF1: 58.7 -> 59.4Improved
use log-probability instead of probability for information gain (ablation)IGPO on Qwen2.5-7B-Instruct · 7 QA datasets avgF1: 59.4 -> 60.2Improved
use log-probability instead of probability for information gain (ablation)IGPO on Qwen2.5-3B-Instruct · 7 QA datasets avgF1: 48.0 -> 48.9Improved
IGPO instead of GRPOQwen2.5-7B-Instruct · MusiqueF1: 25.0 -> 32.7Improved
IGPO instead of GRPO/PPO/GSPOQwen2.5-7B-Instruct · NQF1: 40.3 -> 46.4Improved
IGPO instead of GiGPO step rewardQwen2.5-7B-Instruct · 7 QA datasets avgF1: 47.2 -> 60.2Improved
IGPO instead of GRPOQwen2.5-7B-Instruct · NQ vs GRPOF1No clear effect
information gain reward adds computationIGPO vs GRPO · training step latencyoverhead: <0.02% end-to-endNo 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