INDEX

Explanations

JSON/code syntax elements

The neuron is strongly activated by quotation‐mark tokens, effectively flagging the boundaries of quoted string literals.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

!}

0.61

""}

0.57

()}

0.55

}/>

0.53

^{*}}

0.52

！）

0.52

}{}

0.51

*}

0.51

чками

0.50

();}

0.49

POSITIVE LOGITS

↵↵

0.56

 shew

0.46

(\$

0.43

･

0.42

 sied

0.40

 थीं

0.40

是要

0.39

 SILVER

0.39

</h5>

0.39

 perceive

0.39

Activations Density 0.000%