INDEX

Explanations

The neuron fires on lines that are framed as prompts or directives—specifically on list-style questions and on imperative advice sentences.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 यही

1.34

 stockbild

1.23

.}\

1.06

 سلاټ

0.91

[*][[

0.91

ஃ

0.91

》《

0.87

<0x19>

0.82

<0x04>

0.81

 />}/>

0.80

POSITIVE LOGITS

↵↵

4.38

↵↵↵

1.63

↵↵↵↵

1.41

↵

1.08

‬

1.04

↵↵↵↵↵

0.98

</h2>

0.97

,</

0.84

⠀

0.83

</h3>

0.81

Activations Density 2.504%