Jianwei Ma

Benchmark accuracy is dangerously blind to catastrophic industrial failures: whether an AI destabilizes a power grid or hallucinates subsurface geology, the breakdown traces back to the exact same structural mismatch between physical constraints, data representations, and encoded human intent.

By actively pruning erroneous predictions from its memory graph, HGR achieves a 4.5x reduction in revisits to incorrect regions, demonstrating that retracting hypotheses is as important as generating them for long-horizon embodied navigation.

Personalized federated learning gets a Bayesian upgrade: pFedGM uses Gaussian generative modeling to capture client heterogeneity and outperforms existing methods in diverse scenarios.

RL's inherent resilience to catastrophic forgetting can be harnessed to improve continual learning in GUI agents, outperforming SFT alone.