Winogrande — LLM 评估与基准测试 Dataset

44k commonsense reasoning problems inspired by Winograd Schema challenges. Tests whether models can resolve pronoun ambiguity using real-world knowledge. Requires understanding of causality and physical common sense that pure language statistics can't capture. Included in the Open LLM Leaderboard.

Dataset Details

from datasets import load_dataset
ds = load_dataset("allenai/winogrande")

相关数据集

• MMLU (Massive Multitask Language Understanding) — Benchmarking general knowledge - never train on it
• MT-Bench — Judging multi-turn chat quality with LLM-as-judge
• HumanEval — Benchmarking Python code generation - never train on it
• SWE-bench — Benchmarking real-world software engineering - never train on it

常见问题

← 全部数据集 | Fine-Tuning Guide