Agent Memory Leaderboard releases Cycle 2 results across three tracks
The category-level breakdown matters more than overall rank: memory governance is relatively solved at the top, but personalisation remains weak for every leading system, so teams should not assume a strong memory stack covers user-adaptive behaviour. Enterprise teams building memory-dependent agents, and practitioners in Singapore and ASEAN — where Shopee and Alibaba Cloud DMS directly compete — can use the three tracks to benchmark their own stacks against public baselines.