INDEX

Explanations

`in` followed by location or subject

The neuron is tracking how far into the document each token is (i.e. it encodes the absolute token position from the start).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 adanya

0.71

 primeros

0.68

 operands

0.68

Datei

0.66

 divertida

0.65

 través

0.64

۔

0.64

 através

0.64

พาะ

0.64

 afgelopen

0.64

POSITIVE LOGITS

 which

0.97

which

0.91

 whose

0.84

 here

0.84

here

0.79

 will

0.74

whose

0.72

 enshrined

0.70

in

0.67

is

0.67

Activations Density 0.081%