INDEX

Explanations

code comments

This neuron activates on C‐style block comments, especially the opening “/*” comment delimiters.

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

of

-1.88

様々な

-1.59

 является

-1.56

 termé

-1.52

 vétérina

-1.51

瓛

-1.49

 bordados

-1.48

 takiej

-1.48

 tört

-1.47

 prachtige

-1.46

POSITIVE LOGITS

to

2.72

 "..\..\..\

1.74

who

1.74

更何况

1.66

 except

1.60

 where

1.54

理論

1.53

if

1.52

 unless

1.51

Activations Density 0.010%