INDEX

Explanations

This neuron fires on first-person answer sentences, especially those with personal/possessive framing (e.g. “My favorite part…”).

favorite, primary, target

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

hrter

-0.94

zyści

-0.91

<td>

-0.90

 выпол

-0.89

 sánh

-0.88

 ALREADY

-0.88

cobs

-0.86

%\[

-0.86

 нього

-0.84

抢

-0.84

POSITIVE LOGITS

 would

1.57

1.32

 might

1.30

for

1.27

 probably

1.25

any

1.21

 most

1.18

my

1.17

 really

1.17

 could

1.12

Activations Density 0.009%