INDEX

Explanations

for demonstration or context

The neuron responds strongly to formal, document-structure or meta-text terms (e.g. “guide,” “demonstration,” “background,” “purposes,” etc.).

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

 both

-0.92

PRIVACY

-0.88

したり

-0.88

 from

-0.84

 både

-0.84

 дочь

-0.83

理由は

-0.82

的原因

-0.82

很简单

-0.78

ᆫ

-0.77

POSITIVE LOGITS

 rather

1.59

あくまで

1.38

rather

1.28

 chứ

1.20

 лишь

1.16

\[\

1.09

 '-')

1.08

あく

1.06

而不是

1.04

 thôi

1.03

Activations Density 0.050%