When shape similarity lies: size-ratio artifacts in confusable detection Confusable detection pipelines normalise characters to a fixed canvas before measuring similarity. This hides natural size differences. We re-rendered 2,203 scored pairs (TR39 baseline + novel discoveries) at their original sizes and found 254 with width or height ratios above 2x.
3 models · 4 repos · 3 runs each
，这一点在51吃瓜中也有详细论述
Появились подробности об ответных ударах ВС России по УкраинеМинобороны: ВС России нанесли два массированных и шесть групповых ударов
Instead of yielding one chunk per iteration, streams yield Uint8Array[]: arrays of chunks. This amortizes the async overhead across multiple chunks, reducing promise creation and microtask latency in hot paths.
Сайт Роскомнадзора атаковали18:00