INDEX

Explanations

the truth

The neuron detects the very first word of each new text segment (i.e. the initial token right after a <start> boundary).

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

With

-1.64

-1.52

mathrm

-1.52

 modernas

-1.44

quelize

-1.41

Also

-1.40

Both

-1.38

</

-1.37

),

-1.35

larınız

-1.34

POSITIVE LOGITS

to

1.73

is

1.73

 avon

1.64

and

1.63

 lmao

1.52

 thuy

1.47

"",

1.46

 solidar

1.43

 goud

1.40

 dieß

1.40

Activations Density 0.052%