Entities · Models
OLMoE-1B-7B
5 articles tagged with this entity.
-
Does Mixture-of-Experts Actually Help Inference on Consumer and Edge Hardware? An Empirical Study
-
Pattern Selectivity is Not Task-Causal Structure: A Cross-Architecture Mechanistic Study of Composed-Task Circuits in 1B-Class Language Models
-
Regret Pre-training: Bridging Prior and Posterior Views for Enhanced Knowledge Grounding
-
MobileMoE: Scaling On-Device Mixture of Experts
-
Symmetry-Compatible Principle for Optimizer Design: Embeddings, LM Heads, SwiGLU MLPs, and MoE Routers