INDEX

Explanations

fig. references

This neuron detects formal document markers such as numbered figure/section labels and official title abbreviations (e.g., “FIG.,” “Gov.,” “Rep.”).

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

igure

-1.49

 Reverend

-1.48

Figure

-1.45

 Figure

-1.37

Councillor

-1.29

 figure

-1.24

figure

-1.20

 Professor

-1.01

 Doctor

-1.00

 Figures

-1.00

POSITIVE LOGITS

Fig

5.53

Fig

5.28

 Figs

4.28

Figs

3.66

fig

3.13

Eq

2.64

fig

2.61

Eqs

2.48

 рис

2.33

 figs

2.31

Activations Density 0.068%