📘 CN-SarcasmBench
CN-SarcasmBench is a high-quality Chinese benchmark designed to evaluate large language models on sarcasm understanding, classification, and response generation in real-world online conversations. Based on over 1,200 Bilibili comment threads, it offers rich contextual data and multi-level tasks to reveal the performance gap between current models and human-level sarcastic comprehension. Ideal for researchers and developers aiming to improve nuanced language understanding… See the full description on the dataset page:
https://huggingface.co/datasets/Devon018/CN-SarcasmBench.
