INDEX

Explanations

`resources` configuration

The neuron is primarily activated by numeric and code‐like tokens (e.g. digits, citation numbers, punctuation marks, and inline markup symbols).

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

Terminal

0.63

Semi

0.61

Shr

0.60

estra

0.59

Usually

0.59

Estr

0.58

So

0.57

Proposed

0.57

Steady

0.57

POSITIVE LOGITS

|-

1.34

1.17

"{{

1.08

"|

1.04

"|"

1.02

|.

1.02

 |”

0.97

"*"

0.96

$|

0.92

'|

0.92

Activations Density 0.018%