Federated Nested Learning: Collaborative Training of Self-Referential Memories for Test-Time Adaptation

In three linesFedNL reformulates federated learning as a three-level nested optimization system. Embeds Titans-based linear attention to enable zero-shot test-time adaptation without additional training. Experiments on MMLU and long-context benchmarks show competitive performance and constant inference memory.

Summary generated by Claude — human-verified