INDEX

Explanations

mathematical physics notation

This neuron activates on numeric tokens—especially multi-digit numbers (often dates or other numeric values) embedded in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ěli

0.70

Bowling

0.63

Corre

0.60

yellow

0.60

基地

0.60

ama

0.57

raud

0.57

Geme

0.57

Scott

0.56

 प्रणाली

0.56

POSITIVE LOGITS

 menace

0.66

 lucky

0.55

 unlucky

0.54

up

0.54

 implements

0.53

 finances

0.53

$>$

0.53

ிகை

0.52

 hazard

0.52

 profile

0.51

Activations Density 0.057%