INDEX

Explanations

what and know

The neuron detects colloquial rhetorical question cues—high activations on words like “guess what,” “you know what,” “guess who,” and similar “what”/“who” prompts.

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

 using

-1.85

as

-1.61

any

-1.50

 specially

-1.49

 either

-1.49

 eficientes

-1.42

喽

-1.39

 hohe

-1.37

 Käufer

-1.35

 مدن

-1.34

POSITIVE LOGITS

venant

1.59

 сказать

1.57

 thing

1.57

 they

1.54

 sembl

1.48

᾿

1.48

 Wojciech

1.48

«<

1.47

endforeach

1.46

of

1.45

Activations Density 0.008%