Chuyan Zhou

M1 Okazaki Lab, Institute of Science Tokyo

東京科学大学修士1年
Researching on [Alignment of/Agentic/Diffusion/Scaling] LLMs
Feel free to contact me for collaboration
Japanese/English C1, spoken & written
Click for my Curriculum Vitae

Research Interests

Diffusion LLMs: Architecture, Agentic Post-training/Scaling
Agents: Agentic RL infrastructure and algorithms, Harness Design, Recursive Self-Improvement
Alignment (Post-training) & Nonparametric Scaling of LLMs: Reinforcement Finetuning, Process Reward Models & Design; Distillation

Nonzero Experience / Secondary Interests

LLM Efficiency: Looped Transformers, Soft/Latent Token Transformers, Speculative/Jacobi Decoding (interested in dLLM-powered Decoding Algos), Quantization, KV Cache Compression/Pruning
Mechanistic Interpretability: Sparse Autoencoders, Steering, Logit/Jacobian Lens
Combination of Connectionist and Probabilistic/Symbolic Methods for NLP
AI4S: Neural Methodologies for Biology/Physics/Chemistry/…

Publications

GiLT: Augmenting Transformer Language Models with Dependency Graphs

ACL 2026 Main Conference

Tianyu Huang, Yida Zhao, Chuyan Zhou, Kewei Tu [Paper]

History

2026.09.28: Joined Okazaki Lab officially along with formal enrollment.
2026.07.15: Graduated from ShanghaiTech University with a bachelor degree in Computer Science as a distinguished alumnus.
2026.05.28: Received an offer of admission to 東京科学大学大学院情報理工学院情報工学系知能情報コース (Major of AI, Department of CS, School of Computing, ISCT) for M.S. in the Okazaki Laboratory.
2026.04.07: GiLT: Augmenting Transformer Language Models with Dependency Graphs with my participation, was accepted to the ACL 2026 Main Conference.
2026.03.06: Received an offer of admission to the master’s program in 東京大学大学院工学系研究科技術経営戦略学専攻 (International Technology Management, G30-TMI) at UT, with placement in my first-choice Matsuo-Iwasawa Laboratory.
2026.02.13: Received an offer of admission to the master’s program in 東京大学大学院情報理工学系研究科創造情報学専攻 (Creative Informatics at the University of Tokyo), with placement in my first-choice Nakayama Laboratory.
2025.05: Received the Outstanding Student Award and Scholarship from ShanghaiTech University.
2025.01: Completed the GLOBE Program in University of California Berkeley starting from 2024.08, majoring in Computer Science, GPA 4.0/4.0.
2023.10: Won a silver medal (24th place) teamed in Kaggle Bengali.AI Speech Recognition Challenge.
2023.10: Joined Prof. Kewei Tu’s research group at ShanghaiTech University as an undergraduate researcher.

Projects

Training-free Verifiable Process Reward for LLM Reinforcement Finetuning

Thesis · ShanghaiTech University, BIGAI

2025.10 - 2026.7

We designed a training-free verifiable process rewarding scheme based on formalized theorem provers, consisting of:
- Forcing LLMs to generate natural language reasoning steps in Horn clause $p\land q\to r$ form, enabling semantically complete segmentation of the reasoning process;
- A training-free process reward by integrating a theorem prover/solver called a verifier;
- An LLM teacher in-the-loop to formalize NL reasoning into verifiers;
- A GRPO-like advantage normalizing algorithm for critic-free value estimation, but for PRMs.
We base experiments on logical/math QA datasets to train reasoning models, and evaluate and compare the performance by accuracy improvement after RL finetuning with our proposed process reward.

Reconstruction and Re-evaluation of SFCNN for PLA Scoring

COMPSCI 177 Research Project · ShanghaiTech University

2025.4 - 2025.6

(2025.4-6) Constructed a PyTorch implementation of the model architecture, training, and evaluation of SFCNN (Scoring Function 3D Convolutional Neural Network) for protein-ligand binding affinity prediction.
(2025.4-6) Developed a novel benchmark for similar models to allow inference directly on the 3D structure of protein-ligand complexes instead of on decoupled protein and ligand structures.

LLM-powered Lecture Generation

COMPSCI 194-196 Research Project · University of California, Berkeley

2024.8 - 2024.12

(2024.8-9) Independently developed the backend framework of the lecture generation pipeline using FastAPI as a deployable web service. This backend includes asynchronous task execution via multithreading, task management via an API powered by Redis databases, and a metadata system for managing generated data.
(2024.9-10) Worked as the main developer to integrate the respective model components on the backend framework.
(2024.10-11) Developed an additional LLM-powered QA agent based on the backend that interacts with LLMs using a long context of generated lectures and a RAG system to dynamically index grounding sources (e.g., textbooks).