AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

ResT: Reshaping Token-Level Policy Gradients for Tool-Use Large Language Models

Zihan Lin, Xiaohan Wang, Jie Cao, Jiajun Chai, Guojun Yin, Wei Lin, Ran He

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 12 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
use ResT token reweighting instead of GRPOQwen3-8B · BFCL multi-turn overallaccuracy: 36.00 -> 40.13Improved
use ResT instead of GRPOQwen3-14B · BFCL multi-turn overallaccuracy: 38.88 -> 44.25Improved
use ResT instead of GRPOQwen3-4B-2507 · BFCL multi-turn overallaccuracy: 41.62 -> 50.38Improved
apply SFT or GRPO to Qwen3-4B-2507Qwen3-4B-2507 · BFCL multi-turn overallaccuracy: 48.00 -> 43.00 (SFT); 41.62 (GRPO)Hurt
use ResT instead of GRPOQwen3-8B · API-Bank overallaccuracy: 66.50 -> 70.69Improved
use ResT instead of SFT+GRPOQwen3-4B-2507 · API-Bank overallaccuracy: 68.51 -> 68.68No clear effect
add dynamic scaling reward (ablation)ResT · API-Bank Qwen3-8Baccuracy: 64.15 -> 70.69Improved
add gradients for chain-of-thought tokens (ablation)ResT · API-Bank Qwen3-8Baccuracy: 66.33 -> 70.69Improved
add curriculum learning weights (ablation)ResT · API-Bank Qwen3-8Baccuracy: 65.83 -> 70.69Improved
add curriculum learning weights (ablation)ResT · API-Bank Qwen3-4B-2507accuracy: 67.84 -> 68.68Improved
use ResT instead of SFT on Llama-3.2-3BLlama-3.2-3B-Instruct · BFCL liveaccuracy: 63.04 -> 61.53Hurt
use ResT on Llama-3.2-3B-Instruct vs GRPOLlama-3.2-3B-Instruct · BFCL non-live ASTaccuracy: 74.38 -> 83.19Improved

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