INDEX

Explanations

asking for ideas

The neuron strongly activates on “ideas” in question prompts—i.e. when the text is asking “Any ideas…?” (requests for suggestions).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

辚

-1.77

ᐳ

-1.62

probably

-1.60

 regionale

-1.53

 lmao

-1.48

ఽ

-1.47

 alcune

-1.43

鶿

-1.42

 probablement

-1.42

 hanno

-1.41

POSITIVE LOGITS

or

2.39

now

1.96

 during

1.77

 from

1.73

 after

1.67

 when

1.63

 encourage

1.60

 since

1.59

if

1.59

for

1.59

Activations Density 0.029%