Proud to share that our paper on making ML explanations more trustworthy has been accepted to NeurIPS 2025!

Thanks to my supervisor @lars-kai-hansen.bsky.social and to @sparsity.bsky.social. Supported by @novo-nordisk.bsky.social , @aicentre.dk , and DDSA.

Paper: arxiv.org/abs/2505.11210

The title “DINOv3 Layer 18-19 can solve the mental rotation problem” doesn’t sound as catchy.