INDEX

Explanations

metrics, rates, proportions

The neuron is triggered by programming‐style tokens—identifiers, symbols, and punctuation (e.g. underscores, braces, imports, function/variable names)—i.e. code syntax.

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

 ヌード

-1.30

 Meksiko

-1.24

ofed

-1.16

 Ahli

-1.16

RESULTADOS

-1.15

 salvare

-1.15

 UNIVERSAL

-1.15



-1.13

بسم

-1.12

 mence

-1.11

POSITIVE LOGITS

 overall

1.48

 ratio

1.45

 average

1.34

 total

1.29

1.19

1.15

1.11

 Duration

1.09

1.07

Activations Density 0.018%