INDEX

Explanations

that followed by action verbs

This neuron activates on occurrences of the word “that,” picking out the relative‐pronoun/token “that.”

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

 GEORGIA

-1.88

 Halen

-1.77

 thar

-1.77

 tivesse

-1.75

rio

-1.74

憷

-1.72

 pudesse

-1.71

erobic

-1.71

 thai

-1.69

 AUSTIN

-1.67

POSITIVE LOGITS

of

2.61

at

2.17

is

2.17

for

2.11

or

1.98

 then

1.81

 only

1.77

 remain

1.71

 pozost

1.68

 mempunyai

1.67

Activations Density 0.244%