BenchmarkDiscrete Reasoning Over Paragraphs
DROP
A reading comprehension benchmark that requires a model to read English passages and answer questions involving discrete reasoning (such as addition, subtraction, sorting, counting, and comparison). Evaluation metrics include exact match (EM) and F1 score.
Timeline (last 90 days)
Oct 13
In the author's published 25-question small-sample evaluation, ZY-S1's EM on DROP stayed unchanged at 4 compared with the base model, while F1 improved from 18.24 to 19.76.
Unverified50%