INDEX

Explanations

replace advise

The neuron fires specifically on occurrences of the word “replace.”

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

thschild

-1.77

 вторая

-1.73

見ても

-1.63

っちゃった

-1.62

 originalmente

-1.56

}$,

-1.55

 первая

-1.55

boutin

-1.52

上げて

-1.47

 aparentemente

-1.42

POSITIVE LOGITS

的时候

1.59

枼

1.59

 собрать

1.59

1.57

 invece

1.57

</

1.57

1.55

芣

1.43

どの

1.42

 ส์

1.40

Activations Density 0.029%