Announcement
Reasoning From Scratch finale covers reasoning-model distillation
The series creator presents distillation as an efficient alternative to the reinforcement-learning runs covered earlier.
TLDR
The creator of Reasoning From Scratch shares the series’ eighth and final lecture, presenting distillation as an efficient alternative to earlier reinforcement-learning runs. The lecture outline covers soft and hard distillation, preparing reasoning data, training a Qwen3-0.6B base model and evaluating checkpoints on MATH-500. The creator plans to return after a brief break.
Combined views
8.3K
1 Source, first seen ago
