INDEX

Explanations

describing something as difficult

The neuron activates on the adjective “difficult” (and its immediate context), i.e. on passages describing something as challenging or hard.

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

-1.52

 kanilang

-1.52

 susun

-1.46

鲵

-1.45

Dont

-1.45

 permanente

-1.45

簡介

-1.42

 dezelve

-1.39

 pursuant

-1.37

 APPROACH

-1.37

POSITIVE LOGITS

 However

1.84

 difficult

1.73

we

1.67

{~

1.48

1.45

Of

1.43

 Difficult

1.43

but

1.43

 gereken

1.40

の一つ

1.39

Activations Density 0.041%