INDEX

Explanations

code tags and elements

The neuron is triggered by non‐alphanumeric punctuation characters (dots, backslashes, and angle brackets) typically found in code or markup.

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

all

-0.92

<0xA3>

-0.91

傳統

-0.80

を

-0.79

énier

-0.78

eles

-0.77

 Bosque

-0.76

结语

-0.76

 جاده

-0.76

 lilla

-0.75

POSITIVE LOGITS

ת

0.96

に入れ

0.90

に入れる

0.89

adl

0.88

idit

0.85

 mídia

0.83

 dipende

0.82

ごま

0.81

 també

0.80

eaway

0.79

Activations Density 0.007%