Alessandro Favero

Research Associate@University of Cambridge·英国·OpenReview

研究方向

Theory and science of deep learning · Scaling laws · Statistical physics · Generative models · Diffusion models · Multimodal models · Vision-language models · Model merging · Model editing

4.0

Bigger Isn’t Always Memorizing: Early Stopping Overparameterized Diffusion Models

ICLR 2026Rejected

一作

3.0

Backdoor Unlearning By Linear Task Decomposition

ICLR 2026Rejected

二作

7.3

MEMOIR: Lifelong Model Editing with Minimal Overwrite and Informed Retention for LLMs

NeurIPS 2025Poster

6.6

How Compositional Generalization and Creativity Improve as Diffusion Models are Trained

ICML 2025Poster

一作

6.5

Probing the Latent Hierarchical Structure of Data via Diffusion Models

ICLR 2025Poster

二作

5.8

LiNeS: Post-training Layer Scaling Prevents Forgetting and Enhances Model Merging

合作者 (12)