INDEX

Explanations

substitute, omit, or use

This neuron detects the prose-like, first-person commentary segments (e.g. “I did not add the butter and used non-fat cottage cheese…”) as opposed to ingredient lists or technical 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

 правительства

-0.96

стма

-0.93

 journalistes

-0.91

 beets

-0.90

きましょう

-0.89

 огур

-0.88

 azimuth

-0.88

🧴

-0.88

ň

-0.88

verband

-0.87

POSITIVE LOGITS

 substitute

1.79

 substituted

1.73

 substituting

1.55

omit

1.51

use

1.48

 omitted

1.47

 omit

1.47

 substitutes

1.42

substituted

1.36

subs

1.35

Activations Density 0.032%