INDEX

Explanations

understand and deduce

This neuron strongly activates on instances of the phrase “figure … out” (and closely related “trying to figure out”) indicating it’s detecting the concept of working something out.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 regarding

0.57

 konuda

0.57

ᅯ

0.56

 about

0.54

 पोली

0.54

 शुभेच्छा

0.54

 manfaat

0.53

ğlu

0.52

ummy

0.51

 Opportun

0.51

POSITIVE LOGITS

life

0.58

mart

0.54

ljed

0.52

streaming

0.50

drops

0.50

politics

0.50

 زندگی

0.49

中です

0.49

 tricks

0.49

Life

0.49

Activations Density 0.079%