INDEX

Explanations

positions and limits

The neuron activates on fragments of programming code—particularly Java‐style import/package lines, type names, numeric literals, and other code tokens.

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

wantErr

-0.79

Pfe

-0.78

PMS

-0.77

[]*

-0.73

tlement

-0.73

portun

-0.73

PMS

-0.73

iency

-0.72

offsetY

-0.72

ahue

-0.71

POSITIVE LOGITS

 limit

1.41

 position

1.32

 capacity

1.21

 remaining

1.16

 limits

1.14

limit

1.09

position

1.06

LIMIT

1.05

remaining

1.05

 Limits

1.05

Activations Density 0.007%