INDEX

Explanations

end-to-end

The neuron consistently activates on tokens beginning with or containing the sequence “end,” such as end, ending, endodontic, enduring, endophyte, 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

"><

0.65

`<

0.61

 acyl

0.58

 brows

0.57

 Agri

0.57

 flocks

0.55

 Conditioner

0.55

 attentively

0.55

 crews

0.54

 ří

0.54

POSITIVE LOGITS

End

1.09

End

1.03

end

1.00

END

0.98

orsement

0.93

ToEnd

0.93

 Endo

0.90

端

0.89

owment

0.78

end

0.78

Activations Density 0.132%