INDEX

Explanations

conjunctions and independent words

The neuron is essentially flagging “salient” or high‐information tokens – the document’s key content words (those that stand out, e.g. rare or emphasized terms).

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

大脑

-1.35



-1.27

——–

-1.26

"",

-1.23

 Leur

-1.20

 mereka

-1.19

quele

-1.15

 their

-1.13

Stadium

-1.13

POSITIVE LOGITS

𝙽

1.57

̣i

1.48

 existentes

1.43

ãng

1.43

 системы

1.38

 Mungkin

1.37

 (!__

1.37

.”

1.35

 щодо

1.35

ertia

1.34

Activations Density 0.171%