INDEX

Explanations

yaml code blocks

The neuron is mainly detecting tokens that kick off reference/citation or other metadata‐like fragments (i.e. the non-narrative, structural bits rather than the running prose).

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

ona

0.85

 largement

0.85

ar

0.81

ic

0.80

يت

0.80

 espèces

0.79

 exactement

0.77

 reconnaître

0.75

 traités

0.75

阚

0.75

POSITIVE LOGITS

 stretchy

0.82

 Cardiovascular

0.81

<0xC2>

0.80

 несо

0.77

 به‌

0.77

 Latex

0.77

ПР

0.75

 kitt

0.75

 Есть

0.75

 Новый

0.74

Activations Density 0.000%