INDEX

Explanations

should or can follow

The neuron strongly activates on standalone numerical tokens (e.g. counts, percentages, or other numeric values).

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

給你

0.55

 नसते

0.55

 ለመ

0.54

becomes

0.53

あなたの

0.51

jū

0.50

 lehető

0.50

 ہوتی

0.50

 紹介

0.50

売り

0.49

POSITIVE LOGITS

 should

1.35

should

1.27

 please

1.20

 hãy

1.18

please

1.14

 Should

1.09

 উচিত

1.08

 needn

1.07

请

1.04

 Please

1.03

Activations Density 0.355%