Assistant Professor, EPFL - EPF Lausanne
2 papers at NeurIPS 2025
Aligning pretrained unimodal models with the proposed framework using limited paired data yields ~52% gains in cross-modality zero-shot classification and ~92% in retrieval.
This paper introduces a novel scenario, weak-to-strong generalization under distribution shifts, and proposes Robust AdaptiVe wEightiNg (RAVEN) to tackle this challenge.
