TM
Thinking Machines - Connectionism
Summary
Thinking Machines - Connectionism covering Low-Rank Adaptation (LoRA), Parameter Efficient Fine-Tuning (PEFT), Full Fine-Tuning (FullFT), and Reinforcement Learning (RL). Notable guests include John Schulman.
1episodes
16total claims
12topics covered
1 episodes
LoRA Without Regret
›Sep 29, 2025
Low-rank adaptation (LoRA) can match the performance and sample efficiency of full fine-tuning (FullFT) in most post-training scenarios, provided key hyperparameters and architectural choices are c...
Low-Rank Adaptation (LoRA)Parameter Efficient Fine-Tuning (PEFT)Full Fine-Tuning (FullFT)Reinforcement Learning (RL)+13 more