A unified measurement pipeline on 27 LLM checkpoints shows activation maxima spanning four orders of magnitude, with MoE models 14-23x lower than matched dense models and residual streams carrying the global max in most cases.
When attention sink emerges in language models: An empirical view
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Measuring Maximum Activations in Open Large Language Models
A unified measurement pipeline on 27 LLM checkpoints shows activation maxima spanning four orders of magnitude, with MoE models 14-23x lower than matched dense models and residual streams carrying the global max in most cases.