INDEX

Explanations

H followed by numbers or tech terms

The neuron activates on the “b” token used as the binary‐literal indicator in code (e.g. the ‘b’ in 0b1001).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

and

-1.69

-1.68

</h4>

-1.67

was

-1.62

 детская

-1.57

erobic

-1.57

"]

-1.56

内置

-1.55

 ещё

-1.55

**:

-1.55

POSITIVE LOGITS

↵

2.02

 scattata

1.98

 sağlay

1.86

 kaplama

1.80

 剧照

1.77

似乎

1.70

ientras

1.69

 akhirnya

1.62

 сьогодні

1.60

この

1.59

Activations Density 0.001%