INDEX

Explanations

for convenience, preservation, thought, reach

The neuron primarily activates on numeric tokens (counts, dates, amounts, ages, etc.) in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

1.17

1.12

 becomes

1.05

 youll

0.96

 global

0.94

 transforms

0.88

 will

0.87

0.85

 scalable

0.85

POSITIVE LOGITS

 allegedly

1.34

 wrongdoing

1.28

 якобы

1.28

 उसने

1.27

 fraudulently

1.24

 είχε

1.23

 miała

1.21

 कथित

1.18

 unlawfully

1.18

 قتل

1.17

Activations Density 1.377%