| title | Multiscale Policy Optimization for Long-Horizon LLM Agents |
|---|---|
| authors | Tao Ren, Hui Yang, Jia Deng, Tingyang Chen, Qiyao Wang, Shuo Tang, Ziyang Zeng, Yuhua Jiang, Guanghao Li, Huatong Song, Yining Li, Yifan Luo, Kunbao Ding, Feng Chang, Chuan Hao, Ran Tao, Bryan Dai, Jiafeng Liang, Jinyang Jiang |
| date | 2026-10-26 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.35013 |
| affiliation | IQuest |
| upvotes | 30 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Long-horizon LLM agents interact with their environments through complete assistant turns while generating each action token by token. We observe that critic values vary primarily across turns, motivating optimization anchored at turns with finer credit allocation among tokens. We propose Multiscale Policy Optimization (MPO), combining Multiscale Advantage Estimation (MAE) and turn-level importance sampling to align credit assignment and policy updates at the turn scale. Extensive experiments across search, tool-use, and automation tasks with 4B and 30B-A3B models demonstrate consistent gains over strong RL baselines. |
cite
@misc{ren2026multiscale,
title={Multiscale Policy Optimization for Long-Horizon LLM Agents},
author={Tao Ren and Hui Yang and Jia Deng and Tingyang Chen and Qiyao Wang and Shuo Tang and Ziyang Zeng and Yuhua Jiang and Guanghao Li and Huatong Song and Yining Li and Yifan Luo and Kunbao Ding and Feng Chang and Chuan Hao and Ran Tao and Bryan Dai and Jiafeng Liang and Jinyang Jiang},
year={2026},
eprint={2610.35013},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.35013}
}