INDEX

Explanations

The neuron is highly sensitive to site‐navigation and “call‐to‐action” header tokens (things like “Read More,” “More information,” section headings, ellipses, etc.) rather than ordinary body text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

to

-1.84

on

-1.46

 when

-1.34

[]={

-1.26

 while

-1.26

 taking

-1.25

 puramente

-1.24

 there

-1.23

but

-1.21

५

-1.21

POSITIVE LOGITS

回来了

1.46

 nimmer

1.38

锃

1.35

 verhind

1.34

 familien

1.33

 sugger

1.30

adalupe

1.29

abinieri

1.27

VarChar

1.27

 bewegen

1.26

Activations Density 0.043%