INDEX

Explanations

This neuron fires on imperative‐style instructions or commands (i.e. verbs used to give directions).

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

0.52

 stockbild

0.46

<unused4>

0.46

<unused53>

0.43

‌های

0.43

!”,

0.43

žila

0.42

<unused65>

0.42

!”.

0.42

<unused1>

0.41

POSITIVE LOGITS

↵↵

1.39

↵↵↵

1.17

↵↵↵↵

1.12

↵↵↵↵↵

0.98

↵

0.98

↵↵↵↵↵↵

0.94

<eos>

0.89

↵↵↵↵↵↵↵↵

0.88

↵↵↵↵↵↵↵

0.87

↵↵↵↵↵↵↵↵↵↵↵↵

0.85

Activations Density 2.932%