INDEX

Explanations

inferences and assumptions

The neuron fires on logical‐inference or hedging verbs—words like “implies,” “assumes,” and “suggests.”

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

 still

-1.82

又被

-1.43

も良い

-1.39

-1.34

now

-1.33

憷

-1.32

ﻨ

-1.31

 never

-1.30

 mainstay

-1.30

„

-1.30

POSITIVE LOGITS

 wichtigste

1.83

1.77

1.69

 that

1.67

一定

1.66

鱘

1.57

その後

1.44

知道

1.44

 HAVE

1.42

return

1.41

Activations Density 0.051%