INDEX

Explanations

The neuron is effectively acting as a position counter, firing in proportion to a token’s absolute position (index) in the document.

as/of/by comparisons and relations

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

corrhi

-1.16

 obyvateľov

-1.12

tivate

-1.07

inition

-1.05

},

-1.02

-/

-1.01

çait

-1.01

DEN

-1.00

JAN

-1.00

phant

-1.00

POSITIVE LOGITS

 also

1.51

 those

1.37

 other

1.37

as

1.34

 only

1.27

 both

1.27

 these

1.23

 higher

1.16

 have

1.16

for

1.14

Activations Density 0.121%