INDEX

Explanations

best [x]

The neuron detects superlative recommendation phrases, especially “best X” constructions (i.e. assertions that something is “the best” of its kind).

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

SAME

-1.23

 ſame

-1.20

囖

-1.20

郿

-1.19

 japonaise

-1.19

ṉ

-1.16

umers

-1.14

OOTDTY

-1.13

㌧

-1.11

POSITIVE LOGITS

for

1.53

out

1.30

to

1.27

 money

1.18

 best

1.16

 people

1.15

 ever

1.13

 choice

1.13

According

1.09

 combination

1.08

Activations Density 0.039%