INDEX

Explanations

reporting speech or questions

The neuron strongly detects reporting verbs and speech‐attribution language (e.g. “said,” “tell,” “ask”).

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

↵↵↵

-1.38

季节

-1.30

stücke

-1.30

絝

-1.27

在她

-1.27

erm

-1.20

Hk

-1.20

↵↵↵↵↵↵↵↵↵↵↵

-1.20

↵↵↵↵

-1.20

POSITIVE LOGITS

 their

1.63

『

1.48

 which

1.43

 mismo

1.26



1.24

ೇ

1.23

1.22

 same

1.20

 medes

1.20

 after

1.19

Activations Density 0.022%