INDEX

Explanations

factorial

This neuron is firing on particularly content-heavy or rare tokens—i.e. standout keywords rather than common function words.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

Wenn

0.84

Goed

0.80

Bought

0.77

Codigo

0.76

Laptop

0.76

Restaurant

0.75

Мак

0.75

উদাহরণ

0.75

Restaurants

0.75

تی

0.74

POSITIVE LOGITS

、

1.56

的

1.55

、

1.46

的

1.34

의

1.18

를

1.14

の

1.13

 という

1.08

 का

1.06

を

1.06

Activations Density 0.047%