Evaluating Long-Term Memory for AI Agents: AML Cycle 2 Is Now Open (twitter.com)

The Agent Memory Challenge 2026 Cycle 2 has officially launched, focusing on enhancing long-term memory capabilities in AI agents. This cycle emphasizes the importance of not just accumulating historical data, but effectively retrieving relevant evidence, adapting to changes, and avoiding outdated context in decision-making. The challenge features three tracks—Textual, Coding, and Multimodal—and encourages participation from both open-source methods and commercial products, with a substantial prize pool of over USD 22,000 allocated for eligible open-source teams. This initiative is significant for the AI/ML community as it aims to push the boundaries of how AI agents utilize memory for improved performance and decision-making. By providing a shared Add/Search interface and standardized evaluation metrics, the challenge fosters public engagement and drives innovation, ensuring that results are comparable and transparent. Participants are encouraged to register and contribute to advancing the field of AI memory, potentially leading to breakthroughs in agent efficiency and adaptability.