IC-526GPT-2 small's first token position has residual stream norms more than an order of magnitude larger than all other positions

Leo Gao, Tom Dupre la Tour, Henk Tillman, Gabriel Goh, Rajan Troll, Alec Radford, Ilya Sutskever, Jan Leike, Jeffrey Wu

SourceScaling and evaluating sparse autoencoders

The authors measure residual stream norms across context positions in GPT-2 small and find that the first token position has norms more than an order of magnitude larger than all other positions, except at the first and last layers. Despite the first position being easier to reconstruct in terms of normalized MSE, ablating it to the autoencoder reconstruction causes significantly worse downstream loss and KL divergence than ablating other positions. The authors attribute this to the large-norm direction at the first position serving as a crucial attention resting state, consistent with the attention sink phenomenon described by Sun et al. (2024) and Xiao et al. (2023).

Evidence
observational
Key metric
residual stream norm more than an order of magnitude larger than other positions
Caveat
The exception at early layers and the final layer of GPT-2 small is noted; the phenomenon is layer-dependent. The paper uses a context length of 64 tokens, which the authors acknowledge 'is potentially too few tokens to exhibit the most interesting behaviors.'
Model
GPT-2
Concepts
Register tokens
Methods
Sparse autoencoder / Sparse autoencoders / K-sparse autoencoder / Topk sparse autoencoder / Scaling and Evaluating Sparse Autoencoders / Cunningham et al. 2023 (sparse autoencoders) [supporting]
Related work
Sun et al. 2024 (Massive activations in large language models) [context], Xiao et al. 2023 (Efficient streaming language models with attention sinks) [context]
Related findings
IC-525, IC-527
Extraction
automatic-extraction