multimodalfactbullishMultimodal learning by jointly modeling linguistic and acoustic features leads to more reliable cognitive impairment detection than single-modality approachesMachine Learning29 Jul 2026http://arxiv.org/abs/2607.21496v1