LCM-Lab/LongRewardBench
收藏官方服务：
资源简介：
LongRewardBench是一个针对长上下文场景的奖励模型（RMs）评估的基准数据集，支持4K到128K tokens的上下文长度，包含七个关键任务的1900个高质量样本，用于评估长上下文-回应一致性和判断可靠性。
LongRewardBench is a benchmark dataset for evaluating Reward Models (RMs) in long-context scenarios, supporting context lengths from 4K to 128K tokens, and containing 1,900 high-quality samples across seven key tasks for assessing long context-response consistency and judgment reliability.
提供机构：
LCM-Lab
