INDEX

Explanations

identifies the only is/are

The neuron is triggered by common auxiliary or linking verbs (forms of “to be” and “to have”) that head a clause.

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

 également

-1.73

 просто

-1.69

 mycket

-1.66

 januari

-1.63

 גם

-1.63

でも

-1.59

pielen

-1.59

 PUSTAKA

-1.58

 déjà

-1.57

-1.56

POSITIVE LOGITS

 that

3.25

 those

2.94

two

2.28

 their

2.16

one

2.14

 this

1.98

 when

1.96

 some

1.96

for

1.83

 dotyczą

1.76

Activations Density 0.011%