About me
Short Bio
Basic Information
2003-10-06
Hengyang, Hunan, China
Mandarin, English
Education
Zhiyuan Honors Ph.D. Program, X-LANCE Lab, College of Computer Science
Ph.D. Student
B.Eng. in Automation, minor in Electrical Engineering
Qian Xuesen Honor Class
Exchange Student, Berkeley Global Access (BGA), College of Engineering
Research Interests
My research spans reproducible evaluation, speech-language alignment, speech recognition, language and audio agents, and spatial audio.
📊 Reproducible Research & Reliable Evaluation
The SURE series connects reproducible experimentation, agentic evaluation, and semi-structured speech representations for compositional analysis.
🔗 Cross-Modal Speech Alignment
Aligning speech and language through text-only training, controllable CTC simulation, instruction-conditioned adaptation, and text-guided speech editing.
📚 Surveys
Systematizing Speech LLM architectures, training strategies, tasks, and evaluation to map the field and its open challenges.
🎙️ Automatic Speech Recognition
Unified streaming and non-streaming recognition, ASR error correction, and metrics for reliable transcription.
🌍 ASR in Complex Scenarios
Multilingual and multi-speaker ASR, hotword and contextual recognition, multilingual evaluation benchmarks and normalization, and adaptive data sampling.
🤖 Agentic Systems for Language & Research
Executable protocols, evidence-grounded research, and auditable collaboration for language-based agents.
🎧 Voice & Audio Agents
Building agents that reason over speech and audio evidence, coordinate specialist tools, and make their decisions auditable.
🧭 Spatial Audio
Exploring spatial audio perception and understanding, connecting acoustic scenes with spatial context.
Research Experience
AISpeech, Suzhou, Jiangsu
I work on ASR and multimodal alignment methods that connect speech representations with language model reasoning and instruction following.
Shenzhen Research Institute of Big Data, Remote
I explore Speech LLM-based frameworks for speaker-attributed transcription, aiming to improve speaker consistency and controllability in multi-speaker scenarios.
Hi Lab, Xiaohongshu, Shanghai
I study explicit speaker discrimination and implicit speaker selection strategies for multi-speaker understanding, with an eye toward robust speaker identity modeling under real-world conditions.
Open Source Projects
SURE / Open-Bench
I lead a reproducible speech evaluation toolkit spanning evaluation frameworks, interactive demos, and semi-structured speech annotation.
Audio Agents & Speech Systems
Core contributor to Qwen-Audio-Agent (architecture and backend models); contributor to SLAM-LLM (NPU support) and speech-to-speech (CJK punctuation and Chinese TTS token budgeting).
Research Utilities
I lead paper_claw for paper crawling and ps-slm for speech and language modeling utilities.
Publications (Selected)
- indicates equal contribution. See the full list →
Resume
Personal Information
If you are interested in my research directions, feel free to reach out—I am happy to discuss potential collaborations.