INDEX

Explanations

figures and equations

This neuron detects numeric tokens—numbers, years, figure/panel labels, measurements, and similar digit‐based references.

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

Fig

-2.27

Fig

-2.06

 Figs

-1.91

FIG

-1.89

FIG

-1.88

Figs

-1.80

 Sept

-1.73

fig

-1.68

Feb

-1.68

Oct

-1.55

POSITIVE LOGITS

 Figure

5.09

Figure

4.84

 Figures

2.84

 figure

2.44

igure

2.36

Figures

2.27

 Figura

1.80

 figured

1.76

figure

1.73

 Equation

1.60

Activations Density 0.012%