About

Zhiyu Shen

M.S. Student, School of Computer Science and Engineering, Sun Yat-sen University

I am a second-year Master's student at Sun Yat-sen University, advised by Prof. Yanghui Rao. My research interests include long-term memory for LLM agents, multi-hop reasoning, and emotion recognition in conversations, with a focus on reinforcement learning and agentic approaches.

I am actively exploring 2027 full-time opportunities in LLM agents, reinforcement learning, and applied AI systems.

Email GitHub Scholar ACL Anthology

News

2026.06

One collaborative paper (CrossND) accepted to KDD 2026 ADS Track.

2026.04

🎉 HopWeaver accepted to ACL 2026 Main Conference as an Oral presentation; MemBuilder accepted as a Poster.

2026.01

MemBuilder preprint released on arXiv.

2025.08

One paper (CARE) accepted at EMNLP Main 2025.

2025.07

Started internship at Tencent IEG Lightspeed Studios.

2025.05

HopWeaver preprint released on arXiv.

2025.05

One paper (CoE) accepted at ACL Main 2025.

2024.06

Started M.S. at Sun Yat-sen University.

2024.01

One paper (DesignQuizzer) published at CSCW 2024.

2023.08

Started internship at Tencent TEG AI Lab.

EDUCATION

Sun Yat-sen University

M.S. in Computer Science and Technology · 2024 - 2027 (Expected)

Advisor: Prof. Yanghui Rao

Sun Yat-sen University

B.S. in Artificial Intelligence · 2020 - 2024

Undergraduate studies focused on AI foundations, language technologies, and machine learning.

INTERNSHIPS

Applied Research Intern

Tencent IEG Lightspeed Studios

Jul 2025 - Mar 2026

Worked on game AI speech systems, long-term user memory, and spoken intent understanding for Peace Elite.

Applied Research Intern

Tencent TEG AI Lab

Aug 2023 - Jun 2024

Worked on emotion and motion understanding for virtual humans and narrative CG animation systems.

SELECTED PUBLICATIONS

MemBuilder: Reinforcing LLMs for Long-Term Memory Construction via Attributed Dense Rewards

ACL Main 2026 · Poster

Zhiyu Shen, Ziming Wu, Fuming Lai, Shaobing Lian, Yanghui Rao

We introduce MemBuilder, a reinforcement learning framework that trains a 4B-parameter model to orchestrate multi-dimensional memory construction with attributed dense rewards, outperforming Claude 4.5 Sonnet on long-term dialogue benchmarks.

Paper Code

HopWeaver: Synthesizing Authentic Multi-Hop Questions Across Text Corpora

ACL Main 2026 · Oral

Zhiyu Shen, Jiyuan Liu, Yunhe Pang, Yanghui Rao

We propose HopWeaver, the first framework for automatically synthesizing high-quality multi-hop questions from unstructured text corpora, achieving quality on par with or exceeding human-annotated datasets.

Paper Code

CoE: A Clue of Emotion Framework for Emotion Recognition in Conversations

ACL Main 2025

Zhiyu Shen, Yunhe Pang, Yanghui Rao, Jianxing Yu

We introduce a Clue of Emotion (CoE) framework that progressively integrates multi-dimensional conversational clues with a multi-stage auxiliary learning strategy for emotion recognition in conversations.

Paper Code

OTHER PUBLICATIONS

Collaborative and earlier work.

Cross-Source Reasoning-based Correction for Author Name Disambiguation

Fanjin Zhang, Yunhe Pang, Bo Chen, Zhiyu Shen, Yanghui Rao, Evgeny Kharlamov, Jie Tang

KDD 2026 ADS Track

Paper

CARE: A Disagreement Detection Framework with Concept Alignment and Reasoning Enhancement

Jiyuan Liu, Jiatong Song, Yunhe Pang, Zhiyu Shen, Yanghui Rao

EMNLP Main 2025

Paper

DesignQuizzer: A Community-Powered Conversational Agent for Learning Visual Design

Zhenhui Peng, Qingyu Chen, Zhiyu Shen, Xiaojuan Ma, Antti Oulasvirta

CSCW 2024

Paper

Contact

Research, internships, and collaboration.

Actively exploring 2027 full-time opportunities in LLM agents, reinforcement learning, and applied AI systems.