INDEX

Explanations

return statements

The neuron is primarily detecting numeric literal tokens (constants) in code.

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

and

-1.02

by

-1.01

さんを

-0.90

方で

-0.89



-0.85

Ưu

-0.84

 within

-0.83

 Hence

-0.79

くんの

-0.79

is

-0.78

POSITIVE LOGITS

but

1.12

WART

1.05

 раскра

0.96

ahon

0.95

 نق

0.93

恵

0.92

이블

0.91

pip

0.90

ränken

0.90

 αλλά

0.88

Activations Density 0.004%