INDEX

Explanations

new starting points

The neuron fires strongly on the very first content word of each new paragraph or text block.

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

 имеются

0.76

 имеют

0.73

 имеется

0.73

 considerable

0.73

 umumnya

0.73

 обладают

0.72

 apresentam

0.71

 fragmentary

0.71

 бывают

0.70

 отличаются

0.70

POSITIVE LOGITS

Let

0.89

我现在

0.87

Should

0.84

 Should

0.84

 Shouldn

0.82

Let

0.81

我要

0.80

我去

0.79

 থাকুক

0.79

라도

0.78

Activations Density 0.091%