NovelQA

Dataset Card for NovelQA NovelQA is the first benchmark with the average input length over 200K for testing the long-text ability of LLMs. It comprises the texts of 89 novels (among which 28 are copy-write protected and not publicly published) and 2305 question-answer pairs on the details of these novels. NOTICE: The D

Papers using NovelQA (2)

Tagging-Augmented Generation: Assisting Language Models in Finding Intricate Knowledge In Long Contexts2025ELITE: Embedding-Less retrieval with Iterative Text Exploration2025