INDEX

Explanations

be followed by certain words

The neuron is specialized to detect the exact two‐letter uppercase token “BE.”

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

the

-3.36

↵↵

-2.97

-2.94

for

-2.86

-2.77

-2.63

their

-2.59

-2.55

-2.52

ad

-2.50

POSITIVE LOGITS

tığı

2.95

and

2.48

 tajem

2.30

してしまう

2.30

 vanwege

2.27

 Фі

2.27

 加盟

2.27

 farba

2.25

playera

2.25

pgfp

2.25

Activations Density 0.003%