DualStake: Dual-Path Confidence Calibration in Deep Research Agents
It is found that Evidence Confidence (E-Conf), elicited after the final retrieval step, provides a stronger uncertainty signal than Answer Confidence (A-Conf), elicited after answer generation, and that A-Conf is largely shaped by E-Conf.
Yi-Nuo Xu, Yuanqing Liang, Jianjie Cheng et al.
· 0 citations