INDEX

Explanations

breakfast

The neuron activates strongly on meal‐time words—especially “breakfast” (and related meal terms).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 tématu

-2.03

というわけで

-1.91

 ľudí

-1.84

じゃあ

-1.70

快要

-1.60

 człowiek

-1.57

 spô

-1.56

(ﾟ

-1.55

 листь

-1.54

you

-1.53

POSITIVE LOGITS

1.68

 veulent

1.68

1.54

镡

1.50

 doveva

1.42

 هیچ

1.39

 conç

1.38

 paramount

1.36

1.34

Activations Density 0.009%