Thermodynamics of Prediction
·72 words·1 min
Twitter Thread
Loss Scaling in FP16 Training
·179 words·1 min
Phi-1: The Tiny Model Outsmarting Giants
·255 words·2 mins
Phi-nomenal Code Gen: Phi-1-Base Beats August Models
·172 words·1 min
Tiny Weights, Mighty Models: The Mystery of Weight Decay
·308 words·2 mins
Learning Rates Without Fortune Telling
·415 words·2 mins
LLAMA's Math Leap: From 11% to 49% with Scaling
·424 words·2 mins
When Math Gets Abstract: The 6th Grade Shift
·269 words·2 mins
Understanding "Understanding": Quantifying AI Comprehension
·896 words·5 mins
Web Giants: Users Do the Heavy Lifting
·84 words·1 min