INDEX

Explanations

words of thanks

The neuron fires on tokens that start a new message, paragraph, or list item—i.e. the very first word or number at the beginning of a post or line.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

Enjoyed

-1.01

<th>

-0.94

congrats

-0.87

⑹

-0.84

 irgende

-0.84

SBATCH

-0.84

Congratulations

-0.84

ৰ

-0.82

kého

-0.81

congratulations

-0.81

POSITIVE LOGITS

 thanks

2.27

 Thanks

2.25

 thank

2.17

 Thank

2.06

Thanks

1.97

 thanked

1.96

Thank

1.84

thanks

1.77

thank

1.71

 thanking

1.58

Activations Density 0.022%