INDEX

Explanations

int loops

The neuron fires on programming‐style tokens (loop keywords, variable names, function/array indices, etc.), effectively recognizing bits of code.

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

 each

0.93

แต่ละ

0.82

each

0.81

 EACH

0.80

 Each

0.79

 каждого

0.77

 fiecare

0.76

 foreach

0.75

 каждом

0.74

foreach

0.73

POSITIVE LOGITS

<<-

0.60

kanı

0.51

已经被

0.51

 कहना

0.51

NavBar

0.51

 придется

0.50

 מלא

0.50

пански

0.49

 പോക

0.49

回复

0.48

Activations Density 0.000%