INDEX

Explanations

suggesting possibilities

This neuron detects hedging or suggestion language—words and phrases like "maybe," "just," "like," "experiment," "test," and similar cues that introduce advice or tentative statements.

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

 poliuret

-1.28

 新款

-1.28

🎀

-1.16

zyści

-1.14

 水墨

-1.13

жении

-1.12

ﾒ

-1.12

 pengar

-1.10

ïque

-1.09

*</

-1.09

POSITIVE LOGITS

 also

1.14

1.11

 близь

1.10

 even

1.08

 после

1.07

“

1.05

 will

1.02

 junto

1.01

1.00

 seguinte

0.98

Activations Density 0.027%