Marc Lafon

gallop. Adaptation of vision-language models (CLIP) to downstream tasks using local and global prompts.

cliptta. Official implementation of the NeurIPS 2025 paper: CLIPTTA: Robust Contrastive Vision-Language Test-Time Adaptation

heatood. This repo contains the official implementation of Hybrid Energy Based Model in the Feature Space for Out-of-Distribution Detection (ICML'23).