INDEX

Explanations

section titles with numbers {#sec}

The neuron strongly activates on isolated numeric tokens used as section or anchor identifiers (i.e. the numbers inside the `{#…}` labels).

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

one

-0.95

水墨

-0.90

巽

-0.90

 what

-0.84

bangun

-0.81

苦笑

-0.79

絃

-0.78

ذت

-0.78

樽

-0.77

izace

-0.77

POSITIVE LOGITS

ATH

0.85

 dacă

0.84

 bombilla

0.82

Hvis

0.81

 файлов

0.80

 муж

0.79

ドラマ

0.79

 uwagę

0.79

not

0.77

modular

0.77

Activations Density 0.017%