INDEX

Explanations

for the sake of

This neuron detects milestone time-of-occurrence phrases, especially “for the [ordinal] time” constructions (e.g. “for the first time,” “for the second year”).

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

from

0.63

 from

0.59

auri

0.57

에서

0.56

onson

0.56

牛

0.55

OU

0.54

 成分

0.54

ഴി

0.54

SP

0.54

POSITIVE LOGITS

 sake

2.23

 purposes

1.90

 Sake

1.65

来说

1.52

來說

1.40

 Purposes

1.21

而言

1.17

 betterment

1.14

 reasons

1.12

 foreseeable

1.12

Activations Density 0.319%