INDEX

Explanations

starts a phrase "a"

The neuron fires on the “A” tokens that kick off answer blocks, marking the start of an answer in the Q&A format.

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

și

-2.09

of

-2.08

—

-1.91

the

-1.88

и

-1.75

và

-1.74

-1.73

-1.72

et

-1.70

rozco

-1.61

POSITIVE LOGITS

{"{

1.82

 جذ

1.59

襠

1.55

𝕳

1.52

 ganzes

1.51

 consommer

1.50

BrowserModule

1.49

图为

1.49

Conclusiones

1.48

 pratiquer

1.48

Activations Density 0.033%