1
NeedleChain: Measuring Intact Context Comprehension Capability of Large Language Models
长上下文理解不止“大海捞针”，NeedleChain用链式线索检验模型能否真正串起全文信息，为评测提供新视角。
arXiv:2507.22411v3 Announce Type: replace-cross Abstract: Recent reports suggest that LLMs can handle increasingly long contexts. However, many existi…