INDEX

Explanations

description

The neuron strongly activates on closing delimiters (curly‐brace or XML/HTML‐style closing tags) that mark the end of a structured data block.

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

᯾

-1.71

 colocando

-1.66

 acessórios

-1.65

 ótima

-1.63

šnje

-1.59

"../

-1.57

㋛

-1.56

 colorir

-1.56

 dança

-1.55

 obligación

-1.52

POSITIVE LOGITS

Beskrivning

1.91

1.52

1.50

1.46

1.45

ap

1.44

 with

1.44

1.43

for

1.42

1.38

Activations Density 0.003%