INDEX

Explanations

in terms of

The neuron strongly activates on unusually rare or low‐frequency tokens.

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

ek

0.81

éfono

0.80

ta

0.75

exp

0.73

bserv

0.72

elessly

0.71

ect

0.70

fortunately

0.70

드

0.70

おそらく

0.68

POSITIVE LOGITS

 contoh

0.82

 mencionados

0.82

 thym

0.80

 repug

0.80

 erkek

0.79

𝓲

0.78

 ponctués

0.77

 females

0.77

 commencing

0.77

 extrait

0.76

Activations Density 0.000%