INDEX

Explanations

solving problems

The neuron detects words and phrases that express fixing, resolving, or remedying a problem.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

of

-2.03

 แต่

-1.31

但

-1.30

 Lose

-1.24

 других

-1.24

-1.23

couldn

-1.23

 andern

-1.23

atser

-1.22

Whereas

-1.19

POSITIVE LOGITS

၀

1.49

 when

1.43

 while

1.41

 immediately

1.35

讒

1.34

 komment

1.30

 with

1.27

 quickly

1.27

and

1.24

 napisał

1.24

Activations Density 0.058%