arxiv.org
Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities
As model context lengths continue to grow, concerns about whether models effectively use the full context length have persisted. While several carefully designed long-context evaluations have recently...