INDEX

Explanations

conclusion or final output

The neuron detects words that refer to an ending or final stage (e.g., “final,” “last,” “finish,” “thing”).

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

 anisotrop

1.20

 Сер

1.16

 lysed

1.13

 withstood

1.13

 admire

1.06

प्तान

1.06

 Και

1.06

 subsequ

1.05

″

1.04

 Thời

1.03

POSITIVE LOGITS

 berger

1.25

 концов

1.23

ம்

1.09

isers

1.08

rophes

1.06

тику

1.00

ו

0.99

ท้าย

0.99

্ত

0.98

ter

0.97

Activations Density 0.313%