INDEX

Explanations

same as

This neuron primarily detects occurrences of the word “Same,” marking repeated or follow-up references.

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

 inapropiados

-1.90

谙

-1.81

࿊

-1.70

sve

-1.69

 tüm

-1.67

 सभी

-1.66

 aislados

-1.65

'];

-1.63

 asesinato

-1.56

 sucedido

-1.55

POSITIVE LOGITS

1.81

 SOBRE

1.65

zil

1.62

li

1.61

1.60

ur

1.58

1.55

ed

1.54

 kuhusu

1.54

 inoxidable

1.51

Activations Density 0.005%