INDEX

Explanations

start of turn

The neuron strongly fires on four‐digit numeric tokens (publication years or dates) in citation or “Posted on” lines.

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

 Και

0.46

 králov

0.45

 Dieser

0.44

padă

0.44

 Approx

0.43

ਸ

0.42

 타고

0.42

 testAvg

0.41

 Algebraic

0.41

 Offensive

0.41

POSITIVE LOGITS

0.44

 gasolina

0.44

 ured

0.42

0.40

 спокойно

0.39

у

0.39

 dụng

0.39

 flung

0.38

ating

0.38

ened

0.38

Activations Density 0.000%