INDEX

Explanations

a guest, participant, or role

The neuron fires on occurrences of the indefinite article “a” (often in “as a …,” “was a …,” or “be a …” constructions) that introduce a role or noun‐phrase descriptor.

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

quien

-1.35

 former

-1.16

formerly

-1.11

former

-1.03

who

-1.01

Retired

-1.01

who

-0.97

Former

-0.94

 veteran

-0.91

 tentando

-0.89

POSITIVE LOGITS

 guest

1.39

 pall

1.27

 part

1.22

 participant

1.20

 passenger

1.09

 guinea

1.01

 papild

1.00

 penumpang

0.97

 featured

0.97

guest

0.96

Activations Density 0.085%