INDEX

Explanations

money, work, usage, limit, program, wanna, discretion, executable

The neuron picks out the single most “salient” content word in each text snippet—i.e. the key topic word or head noun/verb that best summarizes that segment.

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

 with

-1.27

時間で

-1.05

bernador

-0.94

GIS

-0.91

di

-0.91

 without

-0.90

って思

-0.88

ۦ

-0.86

時間も

-0.86

-0.85

POSITIVE LOGITS

 zaterdag

1.29

!",

1.28

!");

1.20

՛

1.16

!',

1.13

!";

1.12

1.09

!");

1.09

!\

1.08

!!"

1.06

Activations Density 0.038%