INDEX

Explanations

The neuron fires strongly on the very first content token of each text segment, effectively marking the start of a new chunk.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

上线

0.76

نا

0.74

naire

0.66

0.65

 seasoned

0.64

수가

0.63

/∂

0.63

 scarcely

0.63

ная

0.62

 insensible

0.61

POSITIVE LOGITS

 ratione

0.93

 Planting

0.88

 Earnings

0.80

ctxt

0.80

variation

0.79

vari

0.78

adventure

0.78

agricultural

0.77

adorable

0.77

áreas

0.76

Activations Density 36.164%