INDEX

Explanations

The neuron fires on law‐enforcement and crime‐reporting terms (e.g. Police, arrested, operatives, officers, murder, suspicion).

Q:

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 that

-1.41

 doing

-1.30

 trying

-1.30

how

-1.28

 about

-1.24

 anunció

-1.21

 quería

-1.20

 anuncia

-1.19

 proyek

-1.18

for

-1.16

POSITIVE LOGITS

🪙

1.27

引き続き

1.16

ori

1.14

hey

1.10

 FAILED

1.07

 verminder

1.06

again

1.05

 می‌باشد

1.04

those

1.01

 塗装

1.00

Activations Density 0.004%