INDEX

Explanations

media content descriptions

The neuron fires on words that label a work’s format, category, or metadata (e.g. “online,” “film,” “streaming,” “award,” “service,” “team”).

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

ּוֹ

-1.89

 туры

-1.77

on

-1.73

ﻷ

-1.62

惬

-1.61

摂取

-1.56

 very

-1.56

头上

-1.53

 бампер

-1.50

 食事

-1.45

POSITIVE LOGITS

钣

1.81

 juiz

1.77

шње

1.76

胸口

1.67

 иллюстрация

1.66

 BSEB

1.66

小时候

1.64

ὸν

1.62

 avião

1.61

ᾷ

1.60

Activations Density 0.008%