INDEX

Explanations

evaluative adjectives or negations

The neuron detects subjective evaluation language—phrases where the writer expresses an opinion or judgment (e.g. “it’s a bad question per se,” “not awesome or special,” “should not speak,” 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

 both

-1.43

both

-1.37

Both

-1.21

ทั้ง

-1.19

 nejen

-1.18

 This

-1.18

le

-1.18

着自己的

-1.17

无论是

-1.17

providedIn

-1.17

POSITIVE LOGITS

or

1.60

หรือ

1.59

we

1.59

 или

1.46

שים

1.29

我们就

1.26

Mâ

1.20

 kemudian

1.18

 necessarily

1.16

 necesariamente

1.16

Activations Density 0.038%