INDEX

Explanations

possess a

The neuron fires on words expressing ownership—especially occurrences of “owns/owned a” (e.g. “owns a car,” “owned a general store,” “owns a sawmill,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 into

-1.36

-1.26

as

-1.26

 with

-1.25

to

-1.14

 using

-1.09

out

-1.05

on

-1.04

↵↵↵↵

-1.03

 from

-1.02

POSITIVE LOGITS

 sekarang

1.23

 murale

1.23

quences

1.20

 wonderful

1.20

 maintenant

1.19

 alça

1.16

 Conſ

1.15

 vlas

1.13

 fabulous

1.13

 própri

1.13

Activations Density 0.034%