INDEX

Explanations

can do or works fine

The neuron detects educational learning‐objective phrases—especially “I can” or similar capability statements followed by verbs like describe, identify, read, explain, communicate, etc.

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

—

-1.48

as

-1.36

yled

-1.28

 vraiment

-1.27

 diferente

-1.26

mudah

-1.24

 stärker

-1.23

 même

-1.20

you

-1.20

՝

-1.19

POSITIVE LOGITS

是怎么回事

1.36

也不會

1.29

其妙

1.27

也沒有

1.27

還沒

1.26

说出

1.26

也沒

1.18

 хоте

1.16

 three

1.13

才會

1.11

Activations Density 0.114%