Longhui Zhang, Jiahao Wang, Meishan Zhang, GaoXiong Cao, Ensheng Shi, mayuchi, Jun Yu, Honghai LIU, Jing Li, Min Zhang
Designed a novel benchmark. Proposed a new code acceleration metric. Proposed the BITE method using bidirectional tree editing and progressive learning to accelerate code. (2nd Author)
System Report for CCL25-Eval Task 10: Proposed a novel SRAG-MAV framework that enhances fine-grained Chinese hate speech recognition. (1st Author)
Internships
Tencent Youtu Lab
Research Intern | Mar 2026 - Sep 2026
Agent Skill: Improving on-device model performance with Agent Skill.
Terminal Agent: Data synthesis and model training for Terminal Agent (ICLR 2027 Submission).
Agent Evaluation: Building real-world long-horizon interaction benchmarks from authentic user interaction scenarios.
Verilog Large Language Model Technology Cooperation Project
Research Intern | Jun 2025 - Feb 2026
Huawei-HIT Joint Project supported by CCF-Huawei Poplar Fund.
Participated in developing verilog data governance tools, distilling a Verilog code generation model matching industry-leading performance, and creating supporting IDE plugins.