INDEX

Explanations

code

The neuron fires on tokens containing “code” (e.g. variable names or field labels like “code” in code snippets).

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

-2.47

-2.23

會

-1.86

𝐒

-1.70

 одном

-1.70

ﹶ

-1.61

 leben

-1.59

巨大的

-1.59

(?)

-1.57

POSITIVE LOGITS

code

2.36

 Некоторые

2.14

舜

2.05

 ก็

1.93

ếng

1.88

meteries

1.87

𐐪

1.82

⑮

1.78

 Freunden

1.77

[],

1.76

Activations Density 0.095%