INDEX

Explanations

The neuron fires on code identifiers used as property names or keys (e.g. object‐member names and array/dictionary keys).

data fields and identifiers

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

 zemi

-1.03

 nostre

-0.96

などの

-0.92

 السياسة

-0.91

 pareti

-0.91

ματος

-0.90

vollen

-0.89

älä

-0.89

<0xA3>

-0.89

lighed

-0.88

POSITIVE LOGITS

 location

1.24

 name

1.20

 prices

1.09

){

1.03

fee

0.93

zeß

0.92

 price

0.91

 costs

0.90

 tkan

0.87

 Prozent

0.85

Activations Density 0.012%