INDEX

Explanations

Suwa

The neuron fires whenever it sees the exact two‐letter token “wa.”

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

which

-3.56

我这

-2.86

是我的

-2.52

 nebst

-2.50

 rendus

-2.42

啦

-2.42

蚡

-2.42

甞

-2.41

ة

-2.41

↵↵↵↵↵↵↵↵↵↵↵↵

-2.39

POSITIVE LOGITS

 anguish

2.67

是在

2.58

起伏

2.56

rinology

2.50

 heartache

2.36

同伴

2.30

 tangle

2.25

嫆

2.22

2.19

 uproar

2.17

Activations Density 0.004%