INDEX

Explanations

HTML attributes

The neuron fires on HTML tag names and related attribute names (e.g. “input,” “li,” “link,” “radio,” “name,” “data-depth,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

และ

-1.22

 TÉCNICA

-1.09

λους

-1.08

degna

-1.06

PRNewswire

-1.06

佐

-1.06

jalan

-1.06

і

-1.05

 Projektu

-1.05

 mistura

-1.04

POSITIVE LOGITS

for

1.54

as

1.24

at

1.22

 Although

1.16

 same

1.12

in

1.10

 often

1.10

 usually

1.10

 although

1.09

in

1.08

Activations Density 0.009%