INDEX

Explanations

compared to

The neuron fires on very high‐frequency “function” words (common stopwords like the, a, also, that’s, 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

 ส่วน

0.82

↵

0.76

ร

0.74

Leng

0.73

↵↵↵

0.72

 natomiast

0.71

﹝

0.71

Questa

0.70

0.69

iar

0.68

POSITIVE LOGITS

!),

1.16

oretically

1.13

!).

1.02

!!)

1.02

 admittedly

0.96

 übrigens

0.88

mittedly

0.86

!)

0.83

！）

0.83

 طبعا

0.80

Activations Density 0.617%