INDEX

Explanations

code punctuation

The neuron is detecting numeric tokens—especially multi‐digit and decimal numbers.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 suministro

0.58

 assuntos

0.54

ኮ

0.53

 사망

0.52

 abode

0.51

 conformidad

0.51

 unica

0.50

 asuntos

0.50

ንስ

0.50

 presidente

0.49

POSITIVE LOGITS

 Costs

0.78

 COSTS

0.65

 costs

0.65

 coûte

0.60

<0xBE>

0.59

 পড়ুন

0.59

 Wrapped

0.59

 More

0.58

 Read

0.58

costs

0.57

Activations Density 0.065%