Unigram cross-entropy tracks zero-shot scores from pre-training overlap
Across 10 benchmarks, four corpora, and three scales, lower UCE predicted higher accuracy, and document selection and prompt routing also gained.
1 stories, newest first · one card per story
Across 10 benchmarks, four corpora, and three scales, lower UCE predicted higher accuracy, and document selection and prompt routing also gained.
Today 1.5K readers · 2.4K views · 757 reads · all time 189K views