INDEX

Explanations

rights reserved

The neuron flags tokens that look like programming-style syntax—imports, package names, code comments, hex color codes, version numbers, etc.

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

 anunciou

-0.72

пера

-0.71

dha

-0.69

rada

-0.69

 respondeu

-0.68

躙

-0.68

 번째

-0.66

 portare

-0.66

 explicou

-0.65

 OneNote

-0.65

POSITIVE LOGITS

 حوزه

0.70

 spacecraft

0.70

 Sloane

0.69

NV

0.69

 resources

0.69

CU

0.66

 hoch

0.65

 cover

0.64

Sai

0.64

 tuna

0.64

Activations Density 0.061%