INDEX

Explanations

questions asked

The neuron is strongly activated by punctuation tokens (especially commas and periods), effectively detecting sentence or clause boundaries.

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

Says

-1.05

writes

-0.92

 menulis

-0.87

 berge

-0.85

in

-0.84

кість

-0.82

 étude

-0.82

 kini

-0.82

 것이다

-0.82

wrote

-0.81

POSITIVE LOGITS

 asked

5.31

Asked

5.00

 Asked

4.72

asked

3.34

ask

2.72

ถาม

2.61

 asks

2.58

问

2.56

 questioned

2.53

 when

2.52

Activations Density 0.030%