INDEX

Explanations

numbers and punctuation

The neuron fires strongly on tokens that are structured identifier or field‐name labels containing underscores (e.g. metadata keys).

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

 renom

-1.61

'?

-1.55

führt

-1.54



-1.52

!");

-1.51

戗

-1.50

τας

-1.50

\""

-1.49

'?'

-1.49

 intrig

-1.46

POSITIVE LOGITS

 there

1.94

1.85

 близько

1.76

 actually

1.74

 your

1.74

 vaguely

1.64

 двох

1.59

 оригі

1.57

 literally

1.56

 seems

1.55

Activations Density 0.007%