INDEX

Explanations

answer to

The neuron activates on occurrences of the word “answer,” especially where a numeric score follows.

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

antism

-1.55

sibili

-1.42

 silueta

-1.41

ffiche

-1.39

ární

-1.38

рочный

-1.37

 Román

-1.37

mittlung

-1.36

 canadensis

-1.35

専用の

-1.34

POSITIVE LOGITS

ar

1.54

1.47

they

1.45

</h3>

1.36

兩人

1.34

 They

1.31

說道

1.30

 入っ

1.30

他們

1.27

posicao

1.25

Activations Density 0.002%