INDEX

Explanations

The main thing this neuron does is find sequences of digits and contact‐info patterns (e.g. phone numbers).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 możliwości

0.59

 возможности

0.59

 মর্যা

0.58

 lângă

0.58

 Mikolai

0.58

ıml

0.57

 compreender

0.57

yle

0.56

 livros

0.55

듬

0.55

POSITIVE LOGITS

 hotline

0.90

📞

0.87

calling

0.87

call

0.86

 calling

0.85

 toll

0.83

☎

0.82

 call

0.82

打电话

0.79

 bluff

0.79

Activations Density 0.272%

No Known Activations