INDEX

Explanations

appears to be

This neuron detects hedging constructions—especially forms of “appear(ed/ing) to be” (and similar “seem(ed) to be”) in text.

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

Menurutnya

-1.72

be

-1.64

}_{

-1.62

 about

-1.55

].

-1.52

in

-1.51

all

-1.51

POSITIVE LOGITS

烀

1.69

 pourraient

1.62

谖

1.60

เพราะ

1.56

gæ

1.44

Freitag

1.43

 Công

1.42

pantalones

1.41

噲

1.41

褡

1.39

Activations Density 0.018%