INDEX

Explanations

ignore/ignored

The neuron fires on occurrences of the term “ignore” (e.g. annotations, keywords, imports, or config entries) regardless of context.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

</b>

-0.93

 manner

-0.90

户

-0.90

 называ

-0.90

argent

-0.88

-0.86

 grilling

-0.85

 الخارجية

-0.85

何と

-0.84

 everlasting

-0.84

POSITIVE LOGITS

🚧

1.12

 ceremonia

1.11

 tesis

1.05

 Errors

0.98

 Ignore

0.97

蒸

0.96

CURLOPT

0.96

 Differences

0.93

gefühl

0.93

此

0.91

Activations Density 0.010%