Ph.D. Student at USTC

Zhilin Wang

I am a Ph.D. student at the University of Science and Technology of China (USTC), advised by Prof. Yu Cheng (CUHK).

My research focuses on Agents, Language Models, and AI for Music, with an emphasis on reasoning, adaptation, and intelligent tools for creative work.

Updates

2026.10
Music Arrangement Arena is now available as a research preview for evaluating music arrangement agents. Explore the leaderboard and take part in blind listening comparisons.
2026.07
EvoPolicyGym was released on arXiv.

Publications

Agents

arXiv 2026

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

Zhilin Wang, Han Song, Runzhe Zhan, Jusen Du, Jiacheng Chen, Tianle Li, Qingyu Yin, Yulun Wu, Zhennan Shen, Tong Zhu, et al.

Paper

Language Models

COLM 2026

New Skills or Sharper Primitives? A Probabilistic Perspective on the Emergence of Reasoning in RLVR

Zhilin Wang, Yafu Li, Shunkai Zhang, Zhi Wang, Haoran Zhang, Xiaoye Qu, Yu Cheng.
AI4Math @ ICML 2026

Achieving gold-medal-level olympiad reasoning via simple and unified scaling

Yafu Li, Runzhe Zhan, Haoran Zhang, Shunkai Zhang, Yizhuo Li, Zhilin Wang, Jiacheng Chen, Futing Wang, Xuyang Hu, Yuchen Fan, et al.
ACL 2025

Unveiling Attractor Cycles in Large Language Models: A Dynamical Systems View of Successive Paraphrasing

Zhilin Wang, Yafu Li, Jianhao Yan, Yu Cheng, Yue Zhang.
ACL 2024 · Findings

Spotting AI's Touch: Identifying LLM-Paraphrased Spans in Text

Yafu Li*, Zhilin Wang*, Leyang Cui, Wei Bi, Shuming Shi, Yue Zhang. * Equal contribution.

Music

ISMIR 2026

SymphonyGen: 3D Hierarchical Orchestral Generation with Controllable Harmony Skeleton

Xuzheng He, Nan Nan, Zhilin Wang, Ziyue Kang, Zhuoru Mo, Ao Li, Yu Pan, Xiaobing Li, Feng Yu, Xiaohong Guan.
arXiv 2025

Towards an AI Musician: Synthesizing Sheet Music Problems for Musical Reasoning

Zhilin Wang, Zhe Yang, Yun Luo, Yafu Li, Xiaoye Qu, Ziqian Qiao, Haoran Zhang, Runzhe Zhan, Derek F. Wong, Jizhe Zhou, et al.

Paper

More Publications 11 papers

Agents

2026

arXiv

π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows

Haoran Zhang*, Luxin Xu*, Zhilin Wang*, Runquan Gui, Shunkai Zhang, Haodi Lei, Zihao He, Bingsu He, Chicheng Qin, Tong Zhu, et al. * Equal contribution.

Paper

Language Models

2026

ICML

Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Deliberation

Haoran Zhang, Yafu Li, Xuyang Hu, Dongrui Liu, Zhilin Wang, Bo Li, Yu Cheng.

Paper
ICML Oral

Characterizing, Evaluating, and Optimizing Complex Reasoning

Haoran Zhang, Yafu Li, Zhi Wang, Zhilin Wang, Shunkai Zhang, Xiaoye Qu, Yu Cheng.
ACL Findings

From Drafts to Answers: Unlocking LLM Potential via Aggregation Fine-Tuning

Yafu Li*, Zhilin Wang*, Tingchen Fu, Ganqu Cui, Sen Yang, Yu Cheng. * Equal contribution.

Paper
ICML

Evaluating Parameter Efficient Methods for RLVR

Qingyu Yin, Yulun Wu, Zhennan Shen, Sunbowen Li, Zhilin Wang, Yanshu Li, Chak Tou Leong, Jiale Kang, Jinjin Gu.
arXiv

On the Geometry of On-Policy Distillation

Zhennan Shen, Yanshu Li, Qingyu Yin, Chak Tou Leong, Zhilin Wang, Yanxu Chen, Rongduo Han, Sunbowen Lee, Yi R. Fung.

Paper
arXiv

ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatorics

Shunkai Zhang, Haoran Zhang, Yun Luo, Qianjia Cheng, Haodi Lei, Yizhuo Li, Runzhe Zhan, Zhilin Wang, Bangjie Xu, Yucheng Su, et al.

Paper

2025

ACL

Lost in Literalism: How Supervised Training Shapes Translationese in LLMs

Yafu Li, Ronghao Zhang, Zhilin Wang, Huajian Zhang, Leyang Cui, Yongjing Yin, Tong Xiao, Yue Zhang.
ACL Findings

SEE: Continual Fine-tuning with Sequential Ensemble of Experts

Zhilin Wang, Yafu Li, Xiaoye Qu, Yu Cheng.

2024

ACL

Mage: Machine-generated text detection in the wild

Yafu Li, Qintong Li, Leyang Cui, Wei Bi, Zhilin Wang, Longyue Wang, Linyi Yang, Shuming Shi, Yue Zhang.

2023

ACL

Just Like a Human Would, Direct Access to Sarcasm Augmented with Potential Result and Reaction

Changrong Min, Ximing Li, Liang Yang, Zhilin Wang, Bo Xu, Hongfei Lin.

Projects

Research infrastructure and creative tools for agents and music.

Music · Agent evaluation
Research preview

Music Arrangement Arena

An evaluation platform for agents that arrange existing music for different instruments and ensembles. Explore arrangements, compare musical interpretations, and contribute listening preferences through blind evaluation.

Listen & compare

Browse the repertoire and its alternative arrangements, inspect scores, and take part in blind comparisons.

Evaluate & explore

Explore Elo-based leaderboards with separate human and automated judging views, filtered by arrangement type, model, harness, and reasoning effort.

Visit Arena Blind evaluation Leaderboard Methodology
Agents · Policy evolution
v0.3

EvoPolicyGym

Open-source research infrastructure for studying how coding agents improve executable policies through interaction with environments. A unified interface and standardized protocol connect policy development, bounded feedback, and held-out evaluation.

Develop & assess

Let agents revise policy programs under a feedback budget, then evaluate the resulting policies on held-out cases.

Inspect the evidence

Explore environment-specific leaderboards and experiment replays, including Balatro, Crafter, and NetHack, alongside documentation for running and extending the benchmark.

Project Leaderboard Docs Paper Code
Music · Agent tools
Active development

Echos

A headless, API-first digital audio workstation engine for building AI music tools. It gives agents a programmatic foundation for working with music projects, with the long-term goal of supporting the full production workflow from composition to mastering.

Programmatic control

A service-based API and agent toolkit organize project, track, plugin, editing, and routing operations without relying on a graphical DAW.

Extensible audio engine

An event bus separates core project logic from audio backends. The current backend uses Spotify’s Pedalboard; development focuses on core APIs, plugin hosting, and a broader toolset.

Code & quick start Song creation demo

Experience

Research Experience

MiniMax

2026.05 - Present

Intern

Research on agent context compaction and cross-session memory.

Shanghai AI Laboratory

2024.09 - Present

Research Intern · Shanghai, China

Conducting research on large language models and musical reasoning.

Westlake University (WestlakeNLP)

2023.07 - 2023.09 & 2024.06 - 2024.08

Visiting Student · Hangzhou, China

Mentored by Prof. Yue Zhang. Focused on NLP research.

Education

University of Science and Technology of China

2025 - Present

Ph.D. Student

Researching agents, language models, and AI for music.

Jilin University

2021.08 - 2025.06

Undergraduate Student

Major in Computer Science.

Awards

Nanrui Jibao Scholarship

Scholarship

2025
Huawei ICT Competition

Global Grand Prize

2023
Computer Games Competition

First Prize

2023
CCF BDCI Contest

Third Prize

2022

Academic Service

Conference Reviewer

ACL · EMNLP · NAACL · NIPS · ICML · ICLR, etc.

Zhilin Wang

Agents

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

Language Models

New Skills or Sharper Primitives? A Probabilistic Perspective on the Emergence of Reasoning in RLVR

Achieving gold-medal-level olympiad reasoning via simple and unified scaling

Unveiling Attractor Cycles in Large Language Models: A Dynamical Systems View of Successive Paraphrasing

Spotting AI's Touch: Identifying LLM-Paraphrased Spans in Text

Music

SymphonyGen: 3D Hierarchical Orchestral Generation with Controllable Harmony Skeleton

Towards an AI Musician: Synthesizing Sheet Music Problems for Musical Reasoning

Agents

2026

π-Bench: Evaluating Proactive Personal Assistant Agents in Long-Horizon Workflows

Language Models

2026

Reasoning over Boundaries: Enhancing Specification Alignment via Test-time Deliberation

Characterizing, Evaluating, and Optimizing Complex Reasoning

From Drafts to Answers: Unlocking LLM Potential via Aggregation Fine-Tuning

Evaluating Parameter Efficient Methods for RLVR

On the Geometry of On-Policy Distillation

ComBench: A Benchmark for Rigorous Proof Reasoning and Constructive Realization in Olympiad-Level Combinatorics

2025

Lost in Literalism: How Supervised Training Shapes Translationese in LLMs

SEE: Continual Fine-tuning with Sequential Ensemble of Experts

2024

Mage: Machine-generated text detection in the wild

2023

Just Like a Human Would, Direct Access to Sarcasm Augmented with Potential Result and Reaction

Music Arrangement Arena

EvoPolicyGym

Echos

Research Experience

MiniMax

Shanghai AI Laboratory

Westlake University (WestlakeNLP)

Education

University of Science and Technology of China

Jilin University

Awards

Conference Reviewer