cv

Student Researcher

Stanford SALT Lab

Research on human-centered self-improving AI with Prof. Diyi Yang.

• Studying how recursive self-improvement can be directed toward broader, human-centered intelligence.
• Exploring AI systems that learn from analysis and reflection, then help people in meaningful ways.

Research Collaborator

Princeton / AI2 Collaboration

Collaborated with Prof. Mengdi Wang and Shilong Liu on EEVEE.

• Developed a multi-dataset test-time prompt learning framework for self-improving agents.
• Studied router-prompt co-evolution under heterogeneous real-world task streams.

Research Intern

GAIR Lab, Shanghai Jiao Tong University

Research on self-improving AI systems and autonomous discovery with Prof. Pengfei Liu.

• First author of ASI-Evolve, a closed-loop framework for self-improving AI research.
• Core contributor to ASI-Arch (1.1k+ GitHub stars).
• Designed autonomous pipelines executing 1,773 experiments.
• Contributed to state-of-the-art architecture discovery.

Class Monitor

ACM Honors Class

Led the ACM cohort, coordinating schedules and managing administrative tasks.

Head Teaching Assistant

Shanghai Jiao Tong University

Data Structures (CS1951), lectured by Prof. Yong Yu.

Teaching Assistant

Shanghai Jiao Tong University

Programming (CS1953), lectured by Prof. Huiyu Weng.

Bachelor of Science

Shanghai Jiao Tong University

Computer Science (ACM Honors Class)

• Large Language Models (A+)
• Compiler Design (A+)
• Computer Systems (A+)
• Advanced Compiler Design (A+)

Foresight-Sequoia Talent Development Fund

Zhiyuan College

Selected Top 5 students in Zhiyuan College.

Zhiyuan Honors Scholarship

Shanghai Jiao Tong University

Academic Excellence Award (Top 2% in SJTU).

Provincial Second Prize, CPhO

Chinese Physical Society

Chinese Physics Olympiad.

EEVEE: Towards Test-time Prompt Learning in the Real World for Self-Improving Agents

arXiv preprint

Multi-dataset test-time prompt learning framework for LLM agents under heterogeneous real-world task streams.

ASI-Evolve: AI Accelerates AI

arXiv preprint

Agentic framework for self-improving AI research that closes a learn-design-experiment-analyze loop across data, architectures, and learning algorithms.

ASI-Arch: Autonomous Neural Architecture Discovery

arXiv preprint

Autonomous AI system for neural architecture discovery. The project gained 1.1k+ stars on GitHub within a month and conducted 1,773 autonomous experiments.

ASI-Evolve

Agentic framework for recursive self-improvement in AI research.

• Learn-design-experiment-analyze loop
• Cognition base and dedicated analyzer
• Discovery across data, architectures, and learning algorithms

EEVEE

Multi-dataset test-time prompt learning for self-improving agents.

• Router-prompt co-evolution
• Heterogeneous task streams
• Reduced cross-dataset interference

ASI-Arch

Autonomous AI system for neural architecture discovery (1.1k+ GitHub stars).

• 1.1k+ GitHub stars
• 1,773 autonomous experiments
• State-of-the-art architectures discovered

ACore

Comprehensive operating system kernel implementation in Rust.

• Virtual Memory & Page Allocation
• Process Scheduling & Threads
• VFS & Multiple Filesystems
• User/Kernel Isolation

Imxc

High-performance compiler for a C-like language.

• Performance comparable to Clang
• Mem2Reg & Graph Coloring RegAlloc
• Function Inlining & SCCP
• Incremental Compilation

RISC-V CPU

Verilog implementation of RISC-V core with Tomasulo architecture.

• Tomasulo Algorithm (Out-of-Order)
• RV32IC Instruction Set
• FPGA Deployment Capable
• Instruction Cache