Haosheng (Woody) Gan

Computer Science & Applied Mathematics @ USC

pic.jpg

University of Southern California

Los Angeles, CA

woodygan at usc dot edu

Hey! I’m Woody, a fourth-year undergrad at USC✌️.

I do research on NLP and multimodal models for vision and audio. I’m interested in understanding how AI models actually work under the hood and building better ways to evaluate them.

Currently, I’m working with Professor Vatsal Sharan at USC PALMS Lab🌴. I’m also very fortunate to be advised by Professor Willie Neiswanger and Professor Mahdi Soltanolkotabi. I visited Stanford SALT Lab🌲 this past summer, supervised by Professor Diyi Yang. I’m grateful to learn from my amazing PhD mentors Deqing Fu at USC and William Held at Stanford.

When I’m not training models or writing papers, you can find me playing or watching soccer⚽. I play left wing and people say I’m the next Vini Jr—hit me up if your team needs a good player! I also volunteer through VolunteerMatch. Always happy to connect with others who want to give back to the community!

News

Jan 06, 2026 Putting HUMANS first paper finalized for submission! Working hard on improving the preprint version with more experiments and better analysis. 🚀📄
Jan 04, 2026 AudioJudge accepted to EACL 2026! Looking forward to presenting our work in Morocco in late March. 🎉
Dec 02, 2025 Will be at NeurIPS 2025 in San Diego (Dec 2-7) ! If you’re there, please catch me around the conference - always happy to chat about AI research, collaborations, or just grab coffee! ☕✨

Selected Publications

Putting HUMANS first: Efficient LAM Evaluation with Human Preference Alignment
Putting HUMANS first: Efficient LAM Evaluation with Human Preference Alignment
Woody Haosheng Gan, William Held, Diyi Yang
Under Review , 2026

AudioJudge: Understanding what works in large audio model based speech evaluation
AudioJudge: Understanding what works in large audio model based speech evaluation
Potsawee Manakul*, Woody Haosheng Gan*, Michael J. Ryan, Ali Sartaz Khan, Warit Sirichotedumrong, Kunat Pipatanakul, William Held, Diyi Yang
* Equal contribution
Conference of the European Chapter of the Association for Computational Linguistics (EACL) 2026 , 2026

Textual steering vectors can improve visual understanding in multimodal large language models
Textual steering vectors can improve visual understanding in multimodal large language models
Woody Haosheng Gan*, Deqing Fu*, Julian Asilis*, Ollie Liu*, Dani Yogatama, Vatsal Sharan, Robin Jia, Willie Neiswanger
* Equal contribution
arXiv preprint , 2025

ConceptMix++: Leveling the playing field in text-to-image benchmarking via iterative prompt optimization
ConceptMix++: Leveling the playing field in text-to-image benchmarking via iterative prompt optimization
Haosheng Gan, Berk Tinaz, Mohammad Shahab Sepehri, Zalan Fabian, Mahdi Soltanolkotabi
CVPR Workshop on Generative Models for Computer Vision , 2025

Projects

CAVA: Comprehensive Assessment for Voice Assistants
CAVA: Comprehensive Assessment for Voice Assistants
William Held*, Michael J. Ryan*, Aditya Shrivastava, Ali Sartaz Khan, Caleb Ziems, Minzhi Ella Li, Martijn Bartelds, Michael Sun, Tan Li, Woody Haosheng Gan, Diyi Yang
* Equal contribution

Benchmark Release for Large Audio Models in Voice Assistant Tasks

2025