IC-249The GitHub data-refined LLC identifies the induction circuit heads in Pythia-70m by distinguishing previous-token and induction heads from other head types across layers 2 and 3

George Wang, Jesse Hoogland, Stan van Wingerden, Zach Furman, Daniel Murfet

SourceDifferentiation and Specialization of Attention Heads via the Refined Local Learning Coefficient

The authors apply their weight- and data-refined local learning coefficient (wdrllc) to Pythia-70m, a 70M-parameter released transformer. Using the GitHub code distribution as the data refinement, the relative difference between Pile and GitHub LLC values for layer 2 heads reveals a single outlier corresponding to the model's sole previous-token head. For layer 3, the same measure separates two clusters: candidate induction heads and non-induction heads. These identifications are confirmed using the previous-token score and prefix-matching score from Olsson et al. (2022), with head 3:2 noted as an exception to the induction-head cluster.

Evidence
correlational
Caveat
Head 3:2 is an exception, not cleanly classified as an induction head by the prefix score; the analysis is limited to two layers (2 and 3) of the model.
Model
Pythia
Concepts
Depth-dependent structure
Datasets
GitHub (codeparrot) [eval], The Pile [eval]
Related work
Hoogland et al. (2024) [builds-on]
Extraction
automatic-extraction