INDEX

Explanations

not that

The neuron strongly activates on very common short function words and contraction suffixes (e.g. the “’t” in isn’t/can’t) and other high-frequency “filler” words like all, so, be, actually, that, etc.

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

Răspunde

-0.96

Qué

-0.90

มากกว่า

-0.90

 paixão

-0.89

 visão

-0.89

những

-0.87

Những

-0.87

 []).

-0.86

 unum

-0.86

bior

-0.86

POSITIVE LOGITS

 that

5.13

 THAT

3.27

That

2.80

that

2.48

 That

2.45

THAT

2.20

 ذلك

2.02

 thats

1.88

 того

1.87

那

1.63

Activations Density 0.035%