Ru Zhang

EnCAgg proposes a novel robust aggregation method for federated learning that uses reference clients and advanced clustering techniques to accurately filter dynamic model poisoning attacks while minimizing the loss of benign client gradients.

The paper proposes Test-Time Training for Supervised Causal Learning (TTT-SCL), a novel framework that dynamically generates training data aligned with specific test instances to significantly improve the generalization of supervised causal discovery.

This paper analyzes multi-model self-consuming training, showing that while human curation helps individual models, cross-model interactions can degrade long-term alignment by dampening or inverting the positive effects of human input.