AI Paper Atlas

Research map / Large language models / Agents, code and math

Translate Policy to Language: Flow Matching Generated Rewards for LLM Explanations

Xinyi Yang, Liang Zeng, Heng Dong, Chao Yu, Xiaoran Wu, Huazhong Yang, Yu Wang, Milind Tambe, Tonghan Wang

ICLR 2026 · Large language models › Agents, code and math · approach: Long-horizon LLM agents · 6 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 rectified flow reward model instead of using Proxy LLM rewards directly (ablation)Explanation LLM (Llama-3.1-8B-Instruct) · SMACaccuracy: 0.640 -> 0.764Improved
add rectified flow reward model instead of using Proxy LLM rewards directly (ablation)Explanation LLM (Llama-3.1-8B-Instruct) · MMLUaccuracy: 0.703 -> 0.772Improved
add rectified flow reward model instead of using Proxy LLM rewards directly (ablation)Explanation LLM (Llama-3.1-8B-Instruct) · MathQAaccuracy: 0.694 -> 0.804Improved
use sentence-level rewards instead of sparse last-token reward (ablation)Ours · SMACaccuracy: 0.752 -> 0.764Improved
use sentence-level rewards instead of sparse last-token reward (ablation)Ours · MMLUaccuracy: 0.757 -> 0.772Improved
use sentence-level rewards instead of sparse last-token reward (ablation)Ours · MathQAaccuracy: 0.773 -> 0.804Improved
add cross-attention in the rectified flow reward model (ablation)Ours · SMACaccuracy: 0.662 -> 0.764Improved
add cross-attention in the rectified flow reward model (ablation)Ours · MMLUaccuracy: 0.721 -> 0.772Improved
add cross-attention in the rectified flow reward model (ablation)Ours · MathQAaccuracy: 0.728 -> 0.804Improved
use rectified flow rewards instead of PPO with sentence-level reward (ablation)PPO · SMACaccuracy: 0.688 -> 0.764Improved
use Ours instead of SFT on Qwen2.5-7B-Instruct backboneExplanation LLM · SMACaccuracy: 0.594 -> 0.783Improved
apply rejection sampling to PPO baseline (ablation)PPO · SMACaccuracy: 0.653 -> 0.642Hurt

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