Chao Li

Factorized Neural Operators achieve superior physical modeling by separating transient and persistent responses, leading to enhanced accuracy and interpretability.

HarnessX reveals that evolving agent harnesses through execution feedback can outperform traditional model scaling, achieving up to 44% performance gains in specific tasks.

Even with robust training techniques like EOT, a carefully crafted adversarial patch can reliably fool VIS-IR VLMs and transfer across tasks like classification, captioning, and VQA.

Face symmetry and half-face alignment can be combined to achieve state-of-the-art facial expression recognition.

AI isn't just making things more efficient; it's dissolving the very boundaries of firms and markets, turning them into data nodes within AI-governed infrastructure.

Forgetting isn't a bug, it's a feature: selectively pruning LLM agent memories boosts efficiency by 8%, sharpens content quality by 29%, and eliminates security risks entirely.

Forget hand-tuning: CODO automatically compiles efficient FPGA dataflow accelerators, delivering up to 33.8x speedups on DNN models compared to existing frameworks.

Training more vision models can actually *increase* efficiency, thanks to a novel pre-training strategy that leverages knowledge transfer across a "chain" of models.