INDEX

Explanations

increasing with intensity or number

This neuron detects numerical tokens—numbers and numeric measurements embedded in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

Kabel

-1.11

 that

-1.06

 могли

-0.99

Ϣ

-0.99

 maintain

-0.98

 delicada

-0.96

釧

-0.96

two

-0.96

Kamer

-0.96

outros

-0.94

POSITIVE LOGITS

increasing

1.27

 increasingly

1.23

 zuneh

1.13

 häufiger

1.09

越来越多

1.08

Increasing

1.06

rafat

1.05

ARON

1.04

していく

1.04

PLES

1.04

Activations Density 0.061%