Counting-Stars: A Multi-evidence, Position-aware, and Scalable Benchmark for Evaluating Long-Context Large Language Models.

Mingyang Song, Mao Zheng, Xuan Luo

Browse the full COLING paper archive.