INDEX

Explanations

numbering and phrases like "i recognise", "Embodiments"

The neuron fires most strongly on long runs of whitespace or indentation (extended space‐only segments).

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

Также

-1.92

 légale

-1.88

檮

-1.79

 februari

-1.73

 також

-1.69

Specifically

-1.68

 engelsk

-1.66

ctory

-1.65

吅

-1.62

 dites

-1.61

POSITIVE LOGITS

if

2.23

 they

2.09

2.02

 have

1.99

are

1.96

 this

1.93

 where

1.91

 cannot

1.87

 does

1.84

 doesn

1.78

Activations Density 0.006%