INDEX

Explanations

people and professions

This neuron fires on numeric tokens—particularly multi‐digit numbers (e.g. counts, years, IDs) in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 толкова

-1.03

깎

-0.92

 процес

-0.90

rua

-0.89

dersfield

-0.88

違います

-0.86

 别墅

-0.85

들을

-0.84

 Condiciones

-0.84

降り

-0.83

POSITIVE LOGITS

<eos>

2.67

↵↵↵↵↵↵↵↵

1.56

↵↵↵↵↵↵↵

1.54

↵↵↵↵↵↵

1.53

↵↵↵↵↵↵↵↵↵↵↵

1.47

↵↵↵↵

1.46

↵↵↵↵↵↵↵↵↵

1.45

↵↵↵↵↵

1.44

↵↵↵↵↵↵↵↵↵↵

1.42

</code>

1.41

Activations Density 0.016%