INDEX

Explanations

specifies file name

This neuron picks out the single most content-heavy or “salient” word in each snippet (e.g. key action verbs, topics or names) rather than function or filler words.

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

},

0.85

’,

0.84

”,

0.75

’.

0.73

};

0.71

.’

0.70

`,

0.69

।’

0.69

,”

0.69

’।

0.69

POSITIVE LOGITS

 করিয়ে

0.82

 ভাঙ্গ

0.81

 hostels

0.80

 vết

0.80

 contrat

0.79

 mLast

0.77

віт

0.76

 Garlic

0.76

mre

0.76

 ভাঙ্গা

0.75

Activations Density 0.000%