2 papers
cs.CL2026
From Directions to Regions: Decomposing Activations in Language Models via Local Geometry
Or Shafran, Shaked Ronen, Omri Fahn +3
Activation decomposition methods in language models are tightly coupled to geometric assumptions on how concepts are realized in activation space. Existing approaches search for in…
cs.CL2025
Pretrained LLMs Learn Multiple Types of Uncertainty
Roi Cohen, Omri Fahn, Gerard de Melo
Large Language Models are known to capture real-world knowledge, allowing them to excel in many downstream tasks. Despite recent advances, these models are still prone to what are…