INDEX

Explanations

answering or replying

The neuron activates on verbs (and related tokens) used for attributing or reporting speech—e.g. “answered,” “replied,” “wrote,” 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

 Pembangunan

-1.42

!');

-1.23

粂

-1.18

 femininas

-1.17

 bibliothèque

-1.16

 Litecoin

-1.13

呓

-1.13

 vues

-1.12

 الحب

-1.12

you

-1.11

POSITIVE LOGITS

僦

1.36

1.23

enskap

1.19

 комнаты

1.16

see

1.16

culo

1.15

 öfter

1.14

 detalló

1.14

 Antwort

1.13

嘻嘻

1.13

Activations Density 0.034%