INDEX

Explanations

technical terms and phrases

The neuron spikes on uncommon, specialized or “low‐frequency” tokens—i.e. rare jargon or domain-specific terms.

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

 first

-1.76

 only

-1.76

 just

-1.64

 believed

-1.53

 also

-1.53

to

-1.53

 additional

-1.47

 merely

-1.38

if

-1.38

 even

-1.36

POSITIVE LOGITS

我们

1.98

他们

1.95

THEY

1.58

他們

1.55

它们

1.54

这些

1.52

wouldn

1.51

我們

1.48

couldn

1.45

 ſmall

1.41

Activations Density 0.204%