Full Professor, École de technologie supérieure
2 papers at NeurIPS 2025
We show that interval estimation based methods produce better distilled embedders in multi-teacher distillation settings compared to MSE or Cosine base methods.
LT-Soups merges CLIP models fine-tuned on balanced subsets and retrains the classifier on the full dataset, achieving SOTA head/tail accuracy trade-offs across five benchmarks.
