Sirui Liang

😊 Name in Chinese: 梁思睿

📍 Address: Intelligence Building, Zhongguancun East Road 95#, Beijing, China.

📮 Email: liangsirui2024@ia.ac.cn

Github: Github ｜ Google Scholar: Google Scholar

👋 Hi! I am a second-year Ph.D. candidate at the Institute of Automation, University of Chinese Academy of Sciences, currently in a joint training program with Zhongguancun Academy, under the supervision of  Prof. Kang Liu and Prof. Jian Zhao.

I received my Bachelor's degree from Shandong University in 2024.

My research interests lie at the intersection of natural language processing and large language model.

Generally, I’m interested in the following:

• Agent memory;

• Self-evolving LLM agents in the task space;

• Enhancing complex reasoning ability of LLMs through interpretable methods;

Learning How to Remember: A Meta-Cognitive Management Method for Structured and Transferable Agent Memory

Sirui Liang, Pengfei Cao, Jian Zhao, Wenhao Teng, Xiangwen Liao, Jun Zhao, Kang Liu

We proposes a meta-cognitive memory abstraction method MCMA. By learning structural abstract memory representations and organizing reusable hierarchical abstractions, MCMA transforms memories from inflexible storage into a transferable resource.

Preprint. [Paper][Code]

Bias-Restrained Prefix Representation Finetuning for Mathematical Reasoning

Sirui Liang, Pengfei Cao, Jian Zhao, Cong Huang, Jun Zhao, Kang Liu

We propose Bias-REstrained Prefix Representation FineTuning (BREP ReFT), which enhances ReFT’s mathematical reasoning capability.

AAAI 2026. [Paper][Code]

ABSEval: An agent-based framework for script evaluation

Sirui Liang, Baoli Zhang, Jun Zhao, Kang Liu

We introduced a novel script evaluation dataset, MCScript, consisting of more than 1,500 script evaluation tasks and steps, and developed an agent-based script evaluation framework, ABSEval, to collaboratively evaluate scripts generated by LLMs.

EMNLP 2024. [Paper][Code]

ZhuJiu-Knowledge: A Fairer Platform for Evaluating Multiple Knowledge Types in Large Language Models

Pengfan Du, Sirui Liang, Baoli Zhang, Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao

we introduce the ZhuJiu-Knowledge benchmark.

NAACL Demo 2024. [Paper][Leadboard]

Rm-LR: a long-range-based deep learning model for predicting multiple types of RNA modifications

Sirui Liang, Yanxi Zhao, Junru Jin, Jianbo Qiao, Ding Wang, Yu Wang, Leyi Wei

We propose a novel RNA modification prediction method, namely Rm-LR, which leverages a long-range-based deep learning approach to accurately predict multiple types of RNA modifications using RNA sequences only.

Computers in Biology and Medicine. [Paper][Code]

2020 - 2024    School of Software, Shandong University, Bachelor    (Advisor: Prof. Leyi Wei)

2024 - Now    Institute of Automation, University of Chinese Academy of Sciences and Zhongguancun Academy (Supervisor: Prof. Kang Liu and Prof. Jian Zhao )

中国科学院大学优秀学生，2025，中国科学院大学

山东省优秀学生，2023，山东省教育部

山东省政府奖学金，2022，山东省教育部

My hobbies include 📷 photography, 🖌️ drawing, 🧩 Perler beads, ♠️ Guandan (a card game), 🤴Honor of Kings, and 🐤 Stardew Valley.