← all datasets

WorldMemArena

Emerging
1papers using it
2026first seen

WorldMemArena is a benchmark containing 400 multi-session multimodal tasks designed to evaluate multimodal agent memory through an Action-World Interaction Loop, focusing on the agent's ability to manage evolving personal and task states while providing detailed annotations for diagnostic purposes.

Papers using WorldMemArena (1)

WorldMemArena dataset β€” papers, benchmarks & downloads Β· Large Language Models