INDEX

Explanations

eligible for or to

The main thing this neuron does is detect mentions of eligibility (the word “eligible” and its immediate context).

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

we

-2.34

-2.03

 Dinas

-1.95

 необходимости

-1.91

奁

-1.90

 金属

-1.89

The

-1.88

翃

-1.88

菒

-1.87

)</

-1.84

POSITIVE LOGITS

for

2.48

巿

2.20

ൎ

2.11

 này

2.06

 This

2.05

 compleanno

2.03

to

2.03

Before

1.98

,“

1.95

瀝

1.94

Activations Density 0.004%