INDEX

Explanations

thinking

This neuron detects words and short phrases that prompt reflection or consideration—i.e. calls to “reflect on,” “stop and think,” “consider,” “after doing,” etc.

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

盛り

0.53

arczy

0.51

allelujah

0.50

 accusation

0.49

 praises

0.49

வதில்

0.49

 alluvial

0.48

লালের

0.48

 Lausanne

0.48

娱

0.48

POSITIVE LOGITS

思考

2.58

 thinking

2.39

thinking

2.38

 생각해

2.34

 подума

2.31

 düşün

2.30

 THINK

2.28

 pondering

2.27

想想

2.23

考え

2.22

Activations Density 0.536%