INDEX

Explanations

"me" followed by questions

This neuron strongly fires on direct requests phrased with “tell me,” i.e. the phrase “tell me” in questions or imperatives.

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

-1.99

-1.94

 فهي

-1.89

-1.84

-1.83

媄

-1.74

-1.72

-1.66

 åren

-1.62

to

-1.61

POSITIVE LOGITS

 what

2.05

how

2.00

 about

1.99

 Both

1.86

 after

1.83

if

1.76

 which

1.71

 sufficient

1.66

 Every

1.65

 first

1.55

Activations Density 0.014%