Keynote at 2025 ICCV Trustworthy Foundation Models!

I will be giving a keynote talk at the 2025 ICCV Trustworthy Foundation Models (FM) Workshop. I will talk about our recent series of works on uncertainty estimation of LLMs and VLMs:

- MARS: Meaning-Aware Response Scoring for Uncertainty Estimation in Generative LLMs (ACL'24)

- Do Not Design, Learn: A Trainable Scoring Function for Uncertainty Estimation in Generative LLMs (NAACL'25)

- Reconsidering LLM Uncertainty Estimation Methods in the Wild (ACL'25)

- HARMONY: Hidden Activation Representations and Model Output-Aware Uncertainty Estimation for Vision-Language Models