INDEX

Explanations

introducing conditional clauses

The neuron activates on words that introduce situational or conditional contexts—e.g. “where,” “situations,” “conditions,” “cases.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

of

-2.91

in

-2.13

for

-2.06

to

-2.00

 through

-1.95

at

-1.88

 should

-1.73

 from

-1.72

 will

-1.70

 during

-1.70

POSITIVE LOGITS

it

1.82

 comprenant

1.62

 appartenant

1.44

 secado

1.43

wasn

1.41

without

1.40

 mwenye

1.39

跸

1.39

 telles

1.38

pread

1.37

Activations Density 0.071%