answer-matching-evaluation

【开发与工程】Evaluate language models using open-ended answer generation and semantic matching instead of multiple choice, eliminating test-taking shortcuts and achieving near-perfect alignment with human judgment.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准