Localizing Factual Recall in Transformer MLP Layers
A walkthrough of causal tracing and rank-one model editing for understanding how transformers store and retrieve factual associations.
A walkthrough of causal tracing and rank-one model editing for understanding how transformers store and retrieve factual associations.
A long chunk of text.
Exercises every rendering feature: math, code, sidenotes, tables, and dense typography.