INDEX

Explanations

specific examples or definitions

The neuron fires strongly on rare or unusually infrequent subword tokens, i.e. it detects low‐frequency pieces of text.

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

يها

0.88

INTa

0.83

نګه

0.83

 vudd

0.83

ﯿ

0.82

 następnie

0.81

SMTP

0.81

 ไหร่

0.81

نوان

0.80

જ

0.80

POSITIVE LOGITS

).

0.83

0.81

."

0.77

 MATERIALS

0.76

 COMMENTS

0.73

.)

0.71

".

0.70

").

0.70

）.

0.67

）

0.67

Activations Density 0.001%