INDEX

Explanations

specific word pairings

The neuron strongly responds to specialized or standout tokens—especially proper names, acronyms, and other rare or jargon‐style words.

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

who

-1.73

and

-1.49

 vlastnosti

-1.36

墊

-1.34

ПРИ

-1.30

 conect

-1.30

 ALREADY

-1.29

ԝ

-1.27

Какие

-1.26

点此

-1.25

POSITIVE LOGITS

</em>

1.37

1.35

him

1.27

↵

1.23

↵↵↵

1.21

 flawless

1.17

 larger

1.16

say

1.14

 buona

1.14

 malades

1.13

Activations Density 0.236%