INDEX

Explanations

reasons why

The neuron consistently lights up on content-bearing “cue” words that signal key points or argument structure (e.g. tensions, tackling, reasons, depending, likely, will).

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

 können

-1.34

praš

-1.33

)));

-1.26

können

-1.23

 encontraban

-1.21

lllll

-1.20

izability

-1.17

 См

-1.15

 який

-1.14

 pareció

-1.13

POSITIVE LOGITS

 different

1.63

you

1.44

dymyr

1.27

 everyone

1.26

違います

1.24

却又

1.24

っしゃ

1.23

 продолжает

1.22

 разных

1.22

 United

1.21

Activations Density 0.355%