INDEX

Explanations

last or remaining item

This neuron detects words signaling the final or remaining item in a series (e.g. “last,” “final,” or “remaining”).

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

for

-1.55

 where

-1.54

 lainnya

-1.52

ticulture

-1.44

and

-1.38

 póster

-1.37

RRRR

-1.36

🧐

-1.35

⸮

-1.34

 procesador

-1.32

POSITIVE LOGITS

),

1.39

1.29

”,

1.27

 apresent

1.25

It

1.25

[]

1.25

 previously

1.23

bu

1.23

'',

1.23

na

1.23

Activations Density 0.018%