WorldMemArena

WorldMemArena WorldMemArena is a large-scale multimodal memory benchmark designed to evaluate how well AI systems retain, update, and recall information across extended multi-session interactions grounded in real-world scenarios. Dataset Summary Split Samples Sessions Turns Images Memory Points QA Pairs Agent / GUI 203 4,074 16,252 5,755 13,907 10,382 Agent / Embodied 220 3,075 12,087 6,437 12,186 11,788 Lifelong / Project 18 540 14,580 1,601 6,980… See the full description on the dataset page:

Source: Hugging Face Hub (LCZZZZ/WorldMemArena). Metadata imported from the dataset’s Hub tags.