INDEX

Explanations

missing translation notifications

The neuron fires on tokens that include internal hyphens or underscores (e.g. compounds like “switch‐checked,” “index‐272,” “$progress‐bar‐class,” “a_tracks”), i.e. identifiers split by “-” or “_.”

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

 siè

-1.99

 célè

-1.93

 hunne

-1.74

 empê

-1.74

หรือ

-1.74

 éte

-1.73

 inond

-1.72

 permett

-1.71

 améli

-1.70

 régal

-1.70

POSITIVE LOGITS

 first

1.98

 used

1.83

set

1.80

use

1.77

 only

1.73

 method

1.71

1.66

get

1.66

new

1.64

 provide

1.63

Activations Density 0.020%