Agent Memory benchmarks should measure task success, not retrieval alone(x.com)
1 pointby IreneAI3 hours ago1 comment