INDEX

Explanations

placeholder names or coordinates

The neuron activates on the literal sequence “xyz” when it appears as part of an identifier in code.

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

蹀

-2.77

昻

-2.61

骉

-2.47

不断

-2.45

氿

-2.39

は見

-2.31

 angegeben

-2.27

鹂

-2.23

鹚

-2.20

<bos>

-2.19

POSITIVE LOGITS

2.89

2.81

得上

2.36

 graciosas

2.28

可能會

2.27

𓋼

2.17

貔

2.14

there

2.14

 enlarged

2.11

 addition

2.09

Activations Density 0.004%