INDEX

Explanations

preventing negative outcomes

The neuron activates on language describing resistance, durability, or prevention (e.g. “resistant,” “prevent,” “durability,” “resisting,” “not … react”).

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

 Такие

-1.57

But

-1.45

笵

-1.44

 anunciou

-1.41

 about

-1.40

ツヤ

-1.40



-1.39

 houten

-1.38

旒

-1.37

POSITIVE LOGITS

 which

1.66

 creato

1.53

urusan

1.45

nă

1.43

You

1.43

.*;

1.40

Surprisingly

1.39

ֺ

1.38

licante

1.37

șa

1.35

Activations Density 0.027%