INDEX

Explanations

web application, recompile, object, Breaking, Q, Richard, Step

The neuron activates on Chinese-language (non-Latin) tokens.

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

 撮っ

-2.64

𖥧

-2.53

ਲ਼

-2.52

蟥

-2.47

</sub>

-2.47

the

-2.39

 gouden

-2.34

 IBRARY

-2.27

éroport

-2.25

魠

-2.25

POSITIVE LOGITS

2.52

2.34

");

2.16

 dụng

2.11

</h2>

2.09

If

2.08

…..

2.06

了一个

1.98

 asal

1.84

Now

1.83

Activations Density 0.000%