INDEX

Explanations

for the sake of

The neuron strongly activates on capitalized words in titles or headlines, effectively spotting section or article headings.

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

piers

-1.26

 Time

-1.19

 navegando

-1.17

膾

-1.17

 like

-1.15

懇

-1.14

 find

-1.13

いに

-1.13

答え

-1.12

出来て

-1.12

POSITIVE LOGITS

 sake

1.93

mathrm

1.55

mainly

1.46

ensitivity

1.46

pography

1.43

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.41

//$

1.41

 bakteri

1.40

 jadx

1.39

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.35

Activations Density 0.031%