INDEX

Explanations

product features and brands

This neuron activates on the main “label” words that head off sections or bullet‐point items—i.e. the key title or heading words at the start of feature lists and section headers.

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

extremely

-1.82

ebly

-1.79

trying

-1.75

hehe

-1.74

certainly

-1.73

provides

-1.70

allows

-1.69

 pengar

-1.69

tells

-1.66

׳

-1.66

POSITIVE LOGITS

</h2>

3.06

</h3>

2.55

</h1>

2.53

 from

2.05

</h5>

2.03

 EVERY

1.93

 with

1.88

</h4>

1.88

at

1.87

for

1.83

Activations Density 0.144%