INDEX

Explanations

Diagnostic construct validity

The neuron detects module import statements in code, especially calls to require() with relative paths (e.g. "./" or "../").

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

'>=

-1.59

},

-1.59

 ktor

-1.50

itability

-1.43



-1.38

Ook

-1.34

なぜ

-1.34

önen

-1.32

pielen

-1.30

४

-1.30

POSITIVE LOGITS

 that

1.77

by

1.55

 when

1.53

 Then

1.44

we

1.42

 That

1.30

of

1.25

nyata

1.24

</i>

1.22

ırı

1.19

Activations Density 0.001%