Licheng Liu

Pronunciation: Lee-chuhng Lee-oh

I’m a 3rd year Maths student at Imperial College (currently on a Gap year). I am interested in RL, LLMs and ML Theory. I’m currently looking for a Ph.D. in the 26 fall and an intern opportunity.

I am fortunate to be advised by Prof. Manling Li and Prof. Mingda Qiao.

News

Sep 18, 2025	Our work Online Prediction with Limited Selectivity has been accepted to NeurIPS’25 as a Spotlight (Top 3.5%)! Heartfelt thanks to my mentor Prof. Mingda Qiao for his invaluable guidance and continuous support throughout my research and growth!
Jul 09, 2025	Our work UFO has been accepted to AI4Math@ICML25. Grateful to to all my co-authors!
Mar 19, 2025	I’m excited to share that I’m joining NU MLL Lab 🔬, advised by Prof. Manling Li! I mainly work with Zihan Wang and am grateful for his guidance.
May 23, 2024	I’m grateful to be supervised by Prof. Mingda Qiao working on selective prediction in online settings!

Selected Publications

arXiv
A Simple "Try Again" Can Elicit Multi-Turn LLM Reasoning

Licheng Liu, Zihan Wang, Linjie Li, and 5 more authors

2025

Abs arXiv Bib

In this paper, we propose Unary Feedback as Observation (UFO), a simple multi-turn reinforcement learning method that helps large reasoning models reflect on mistakes and improve through minimal feedback like “Let’s try again.” UFO improves multi-turn reasoning accuracy by up to 14% while maintaining single-turn performance, enabling more deliberate and flexible problem solving.
@misc{ufo, title = {A Simple "Try Again" Can Elicit Multi-Turn LLM Reasoning}, author = {Liu, Licheng and Wang, Zihan and Li, Linjie and Xu, Chenwei and Lu, Yiping and Liu, Han and Sil, Avirup and Li, Manling}, year = {2025}, primaryclass = {cs.LG}, }
arXiv
Online Prediction with Limited Selectivity

Licheng Liu and Mingda Qiao

2025

Abs arXiv Bib

In this paper, we introduce a model of Prediction with Limited Selectivity (PLS) where the forecaster can start the prediction only on a subset of the time horizon. We study the optimal prediction error both on an instance-by-instance basis and via an average-case analysis. We introduce a complexity measure that gives instance-dependent bounds on the optimal error. For a randomly-generated PLS instance, these bounds match with high probability.
@misc{PLS, title = {Online Prediction with Limited Selectivity}, author = {Liu, Licheng and Qiao, Mingda}, year = {2025}, primaryclass = {cs.LG}, }

arXiv

RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning

Zihan Wang, Kangrui Wang, Qineng Wang, and 15 more authors

2025

Abs arXiv Bib

@misc{ragen,
  title = {RAGEN: Understanding Self-Evolution in LLM Agents via Multi-Turn Reinforcement Learning},
  author = {Wang, Zihan and Wang, Kangrui and Wang, Qineng and Zhang, Pingyue and Li, Linjie and Yang, Zhengyuan and Jin, Xing and Yu, Kefan and Nguyen, Minh Nhat and Liu, Licheng and Gottlieb, Eli and Lu, Yiping and Cho, Kyunghyun and Wu, Jiajun and Fei-Fei, Li and Wang, Lijuan and Choi, Yejin and Li, Manling},
  year = {2025},
  primaryclass = {cs.LG},
}