INDEX

Explanations

# and numbers

The neuron is detecting URL fragments—tokens that are parts of web addresses (domains, paths, numeric IDs, file extensions, 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

 picnic

-1.43

 picnics

-1.32

Pic

-1.31

 Picnic

-1.19

Pic

-1.07

 their

-1.02

pic

-0.99

for

-0.98

PIC

-0.97

POSITIVE LOGITS

 dekoratif

1.18

 attente

1.13

 abiye

1.08

 obiet

1.07

 peluche

1.06

铙

1.05

Ᏹ

1.03

 vilja

1.02

 Enciclopedia

1.02

 cuivre

1.02

Activations Density 0.002%