The paper introduces InMind, a 125-task, expert-verified benchmark covering ten life domains, with 113 tasks grounded in citable public sources. Its paired controls distinguish missing storage, missing bridging knowledge, and retrieval failure. When decisive memories remain in context, the backbone answers 84.0% of indirect queries. When systems must retrieve those memories, six vector, graph, and agentic memory systems achieve at most 14.4%, despite recalling the same facts on demand at up to 100%. Increasing embedding dimensionality eightfold improves target recall but barely narrows the answer gap.
No heat snapshots are available in the last 24 hours.