INDEX

Explanations

conditional logic or programming constructs

The neuron detects programming comparison and logical‐operator tokens (e.g. !=, &&, =>).

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

最大限

0.66

cath

0.63

 괜찮

0.57

雖然

0.57

 walaupun

0.57

 знают

0.56

 сверх

0.56

 jeszcze

0.56

虽然

0.55

 смогут

0.55

POSITIVE LOGITS

1.22

(!

1.12

!\

1.05

!"

0.98

！

0.96

!(

0.94

！（

0.91

!")

0.88

!_

0.88

!$

0.87

Activations Density 0.003%