INDEX

Explanations

hat, head, of, worn

This neuron detects the character sequence “hat” (e.g. in tokens like “widehat”, “hat_”, “hat”, etc.).

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

-2.22

-2.13

禤

-1.90

鬭

-1.81

 evident

-1.77

 detalj

-1.77

this

-1.75

when

-1.74

縢

-1.74

but

-1.73

POSITIVE LOGITS

2.25

2.16

会在

1.82

 глав

1.73

 illuminate

1.72

--

1.70

をする

1.69

 locations

1.66

 Millions

1.60

of

1.59

Activations Density 0.001%