INDEX

Explanations

choosing a favored candidate

The neuron fires on occurrences of the word “candidate” (i.e. mentions of political candidates).

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

 Inſ

-1.37

瘫

-1.23

 également

-1.22

тельных

-1.21

 bascul

-1.20

 rosse

-1.16

就已经

-1.14

 kahve

-1.14

 bré

-1.14

 edil

-1.13

POSITIVE LOGITS

at

1.61

 whom

1.43

 supported

1.36

who

1.27

 sürekli

1.27

 using

1.27

 doing

1.23

 möge

1.22

 seems

1.21

 ")";

1.20

Activations Density 0.044%