INDEX

Explanations

The judge patient document

This neuron fires on numeric tokens that include a decimal point (i.e. floating-point numbers).

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

妷

0.67

醝

0.63

 できる

0.61

 Амери

0.60

>.</

0.59

 Nucleaires

0.58

ၼ

0.58

 иногда

0.57

 próp

0.56

 propried

0.55

POSITIVE LOGITS

if

0.58

for

0.55

you

0.53

all

0.53

0.52

 holds

0.52

0.50

”,

0.50

 thoughtfully

0.49

 your

0.49

Activations Density 0.000%