Omni-modal assistants must remember not only what users say, but also what they show and convey. Existing long-term memory benchmarks are mostly text-and-ima…
机构:港科大
来源:arXiv 2610.29249 | AI4Papers 论文推荐平台