INDEX

Explanations

energy, fatigue, tiredness

Explanation of neuron 4 behavior: the main thing this neuron does is find descriptive adjectives and adverbs that convey states, qualities, or intensities (for example “dazed,” “tired,” “sluggish,” “weak,” “quickly,” “exhausted”).

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

 idolat

0.59

 promiscu

0.58

不斷

0.57

bul

0.53

 lebens

0.53

入门

0.53

 berhenti

0.53

 egiten

0.52

 screams

0.52

ioctl

0.52

POSITIVE LOGITS

 fatigued

1.94

 tiredness

1.91

 tired

1.80

 fatigue

1.78

 Fatigue

1.77

 Tired

1.65

tired

1.64

疲れ

1.63

疲

1.58

 थकान

1.41

Activations Density 0.455%