Interpretability

Mechanistic Interpretability Workshop at ICML 2024
The Scaling Behavior of Large Language Models
Visualizing Neural Network Imagination
Understanding Addition in Transformers
Increasing Trust in Language Models Through the Reuse of Verified Circuits
What Does GPT Store in Its MLP Weights? A Case Study of Long-Range Dependencies
DeepDecipher: Accessing and Investigating Neuron Activation in Large Language Models
Detecting Edit Failures in Large Language Models: An Improved Specificity Benchmark
The Larger They Are, the Harder They Fail: Language Models Do Not Recognize Identifier Swaps in Python
Neuron to Graph: Interpreting Language Model Neurons at Scale