Across 12 language models, neurons in early-middle layers tend to add the direction they detect (enrichment), while later layers tend to reduce it (depletion), based on input-output weight cosine similarity.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Understanding Gated Neurons in Transformers from Their Input-Output Functionality
Across 12 language models, neurons in early-middle layers tend to add the direction they detect (enrichment), while later layers tend to reduce it (depletion), based on input-output weight cosine similarity.