INDEX

Explanations

align and straight

alignment with goals or values

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

0.66

uğu

0.63

0.61

of

0.60

pl

0.58

ex

0.57

0.56

ية

0.55

 verder

0.54

POSITIVE LOGITS

 Straight

0.95

 straight

0.91

 alignment

0.91

 aligned

0.89

Straight

0.88

 STRAIGHT

0.88

 straightened

0.86

 aligns

0.85

 aligning

0.85

 Alignment

0.84

Activations Density 0.024%