Paper
In-context Learning and Induction Heads
"Induction heads" are attention heads that implement a simple algorithm to complete token sequences like [A][B] ... [A] -> [B]. In this work, we present preliminary and indirect evidence for a hypothesis that induction heads might constitute the mechanism for the majority of all "in-context learning" in large transformer models (i.e. decreasing loss at increasing token indices). We find that induction heads develop at precisely the same point as a sudden sharp increase in in-context learning ability, visible as a bump in the training loss. We present six complementary lines of evidence, arguing that induction heads may be the mechanistic source of general in-context learning in transformer models of any size. For small attention-only models, we present strong, causal evidence; for larger models with MLPs, we present correlational evidence.
Authors: Olsson, Catherine · Elhage, Nelson · Nanda, Neel · Joseph, Nicholas · DasSarma, Nova · Henighan, Tom · Mann, Ben · Askell, Amanda · Bai, Yuntao · Chen, Anna · Conerly, Tom · Drain, Dawn · Ganguli, Deep · Hatfield-Dodds, Zac · Hernandez, Danny · Johnston, Scott · Jones, Andy · Kernion, Jackson · Lovitt, Liane · Ndousse, Kamal · Amodei, Dario · Brown, Tom · Clark, Jack · Kaplan, Jared · McCandlish, Sam · Olah, Chris