INDEX

Explanations

links in markdown

The neuron fires on non‐standard alphanumeric tokens—things like URL fragments, file names, and other random letter-digit sequences rather than ordinary words.

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

/#{

0.63

ának

0.62

/]

0.62

über

0.60

 禁止

0.59

ianza

0.58

ït

0.57

/<%=

0.57

#{

0.57

소를

0.57

POSITIVE LOGITS

コメント

0.56

 генера

0.55

 prior

0.54

 blob

0.54

 commits

0.54

文件中

0.51

 поперед

0.51

blob

0.50

 memer

0.50

런치

0.50

Activations Density 0.027%