INDEX

Explanations

less than

This neuron fires strongly on the last salient content word of each block or paragraph, effectively marking each segment’s concluding key term.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

<unused1741>

0.51

<unused346>

0.50

<unused730>

0.49

<unused299>

0.49

<unused1896>

0.49

<unused1012>

0.48

<unused1748>

0.48

<unused1938>

0.48

<unused320>

0.47

<unused1922>

0.46

POSITIVE LOGITS

0.79

0.75

0.73

0.66

0.65

0.62

0.60

and

0.59

Activations Density 0.000%