Models & Agents
Ep 223: An LLM edited just 0.016 percent of a small model's weights to raise its math benchmark…
10:42
Models & Agents An LLM edited just 0.016 percent of a small model's weights to raise its math benchmark score from 7.9 to 48.1 percent. What You Need to Know: MIT CSAIL researchers demonstrated that frontier LLMs can directly read and rewrite the weights of smaller neural networks using natural language instructions, without gradient descent or retraining. ... Sources: analyticsindiamag.com ·…