INDEX

Explanations

lifecycle methodologies

The neuron strongly activates on relatively uncommon or specialized content words—those that tend to carry high information (e.g. “violating,” “transactions,” “arguing”) rather than on common function 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

hesia

0.73

 画像

0.68

acy

0.67

ди

0.65

可能です

0.65

ী

0.64

ādi

0.64

0.63

ام

0.63

δε

0.61

POSITIVE LOGITS

HANDLER

0.92

 semifinal

0.91

potato

0.90

DTS

0.88

 grout

0.88

ONTO

0.87

 gooey

0.86

 postulate

0.86

 subplot

0.85

StatusCode

0.84

Activations Density 0.000%