OpenNovelty: An Open-domain Benchmark for Evaluating the Open-ended Novelty of Language Models
Published in arXiv preprint, 2026
OpenNovelty asks a difficult question for scientific-intelligence systems: can a language model reason about whether an idea is actually novel in an open-ended setting?
Show citation
Wenqing Jing, Jingyi Deng, Huayu Sha, Jiajun Jiao, Shichun Liu, Changhao Jiang, Ming Zhang, Zhiheng Xi, Shihan Dou, Tao Gui, Qi Zhang, and Xuanjing Huang. 2026. OpenNovelty: An Open-domain Benchmark for Evaluating the Open-ended Novelty of Language Models. arXiv preprint arXiv:2601.01576.