How Language Models Actually Think
The Data Exchange with Ben Lorica
00:00
Towards Debuggers for Models
Emmanuel likens interpretability tools to debuggers and describes efforts to inspect neuron states during execution.
Play episode from 14:14
Transcript
