Papers accepted at ICLR 2026

Emanuele Rodolà

Last updated on Feb 17, 2026

Our five papers "Navigating the Latent Space Dynamics of Neural Models", "Language Models are Injective and Hence Invertible", "MASS: MoErging through Adaptive Subspace Selection", "Implicit Inversion turns CLIP into a Decoder", and "Video Unlearning via Low-Rank Refusal Vector" were accepted at ICLR 2026 main conference!