INDEX

Explanations

good evaluations

This neuron activates strongly on the word “good” (especially in “too good to …” or similar positive‐evaluation phrases).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

躂

-1.42

or

-1.40

-1.38

 desmotiv

-1.33

 Sarkozy

-1.31

 anledning

-1.31

 keinen

-1.30

 geweldig

-1.29

+;

-1.27

meisters

-1.25

POSITIVE LOGITS

andowski

1.60

1.47

philis

1.42

nessy

1.41

 Aussehen

1.41

</h2>

1.39

 Unglück

1.39

 Gewicht

1.37

partly

1.34

brigens

1.34

Activations Density 0.012%