A researcher affiliated with the Paul G. Allen School of Computer Science & Engineering at the University of Washington.
How media typically covers Inna Wanyin Lin
Inna Wanyin Lin as author
A self-training framework for Vision-Language Model judges can achieve competitive performance without human annotations, with a Llama-3.2-11B model outperforming larger models like GPT-4o and Claude 3.5 Sonnet on multimodal evaluation benchmarks.
“Author of "Training a VLM Judge Without Human Labels" on arXiv”