LLM
Dissection
Lab
Model:
Claude 3 Haiku
Claude 3.5 Sonnet
GPT-4o Mini
Llama 3.1 70B
Mistral Large
back to portal
Model Architecture
Weights
Activations
Attention
Gradients
Attention Pattern:
Head 0