news

Jan 26, 2026 Our “Avoid Catastrophic Forgetting with Rank-1 Fisher from Diffusion Models” was accepted to ICLR 2026!
Sep 24, 2025 We showed Diffusion models have a rank-1 Fisher information matrix and achieve no catastrophic forgetting in class incremental image generation task.
Sep 18, 2025 Our “Deep Taxonomic Networks for Unsupervised Hierarchical Prototype Discovery” was accepted to NeurIPS 2025 as a poster. See you in San Diego!