INDEX

Explanations

code assignment and list processing

This neuron fires on interface-style or instructional tokens—i.e. words that serve as action prompts or section cues (like “Add,” “trademarks,” “anytime,” “explain,” “tell,” “end,” “written”).

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

 cierra

0.63

 thổi

0.62

 fleshy

0.61

 Organic

0.61

 persever

0.59

 სი

0.58

aderie

0.58

 tenacity

0.57

 మి

0.57

 lleva

0.56

POSITIVE LOGITS

gpg

0.55

/**

0.54

いない

0.52

Await

0.52

าง

0.51

}".

0.51

"/>.

0.51

INPUT

0.50

VAT

0.50

/>}/>

0.50

Activations Density 0.000%