INDEX

Explanations

achieve some goal

The neuron fires on words that denote carrying out or achieving an action—verbs like “do,” “make,” “accomplish” and nouns like “task,” “purpose,” or “endeavor.”

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

⏑

-1.36

 เรา

-1.27

 takich

-1.24

 overarching

-1.24

 horrid

-1.23

茈

-1.22

を作る

-1.17

 mały

-1.17

 accett

-1.16

 frequentemente

-1.16

POSITIVE LOGITS

 with

2.03

 this

1.77

 because

1.65

 before

1.61

 some

1.57

 most

1.53

 sehr

1.48

 several

1.45

 Several

1.43

but

1.40

Activations Density 0.143%