INDEX

Explanations

exploitation, Japanese AV, harm

The neuron consistently lights up on passages that report quantitative metrics or formal audit‐style figures—i.e. statistical or compliance data references.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 Valves

0.57

 valve

0.53

 piston

0.53

 carburetor

0.52

 Valve

0.51

샷

0.51

Hib

0.51

HIP

0.50

Substring

0.50

 hipert

0.50

POSITIVE LOGITS

 labor

1.19

 labour

1.17

ILO

1.16

 explo

1.14

 শ্রম

1.13

 exploitation

1.08

劳动

1.07

 श्रमिकों

1.04

Labor

1.03

労働

1.03

Activations Density 0.189%