INDEX

Explanations

another instance

The neuron fires on the document’s key topical or technical “keyword” tokens—i.e. the main content words that define the subject (e.g. “load balancing,” “player,” “Ad,” “connections,” etc.).

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

Ooh

-1.00

etiqueta

-0.90

へ向

-0.88

噢

-0.88

רכז

-0.88

assemblée

-0.84

 diversified

-0.84

 faciliter

-0.83

choix

-0.82

 applicazioni

-0.82

POSITIVE LOGITS

 again

1.54

今度は

1.18

อีก

1.16

 with

1.13

 another

1.10

 další

1.05

再度

1.03

再次

1.02

再び

1.02

 other

0.98

Activations Density 0.014%