INDEX

Explanations

explanation of code

The neuron detects meta-commentary cues—words that introduce or signal explanations, clarifications, or illustrative framing (e.g. “note,” “illustration,” “purposes,” “clarification,” “what,” “trying,” etc.).

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

Karakter

-1.38

kapital

-1.35

karakter

-1.22

and

-1.19

ığımız

-1.16

 cannot

-1.15

ığım

-1.11

 ингреди

-1.11

ARROLL

-1.10

Ꮕ

-1.10

POSITIVE LOGITS

 miſ

1.25

 бого

1.24

消失了

1.23

 bribes

1.18

 marta

1.18

 ajustable

1.17

VIH

1.16

国道

1.16

 ikat

1.16

 volverá

1.15

Activations Density 0.070%