INDEX

Explanations

code identifiers followed by special characters

The neuron fires on number tokens (numeric literals or any token containing digits).

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

Mel

-0.93

new

-0.88

and

-0.88

tokio

-0.87

 with

-0.86

autoreleasepool

-0.85

ValuePair

-0.85

 Dostupné

-0.85

 albo

-0.85

strokeWeight

-0.84

POSITIVE LOGITS

striping

1.02

 mußte

1.02

 ersche

0.99

よくある

0.97

コスプレ

0.95

ステンレス

0.95

 ſeveral

0.95

женские

0.95

 ganadores

0.94

 remercier

0.93

Activations Density 0.025%