Wei Li

You can now get real-time (825 FPS) crack detection on UAVs without sacrificing accuracy, thanks to a new attention-guided lightweight CNN.

Multimodal agents can now reason, plan, and execute actions more effectively by integrating perception as a core component, not just an auxiliary interface.

LVLMs hallucinate less when you intervene *before* they start generating, by cleaning up the initial Key-Value cache with modality-aware steering vectors.

One-step image super-resolution can now achieve state-of-the-art realism and fidelity thanks to a novel framework that guides generative flows to avoid trajectory drift and prior collapse.

Multi-level preference alignment in SignDPO significantly reduces semantic drift, outperforming traditional gloss-free models and challenging gloss-based benchmarks.

Surprisingly, deep ensembles *worsen* performance under physics constraints in PINNs for methane sorption, highlighting the importance of uncertainty quantification method selection in physics-informed machine learning.

LLMs can now automatically validate smart contract vulnerabilities by generating and refining profit-driven exploits, significantly reducing the burden of manual verification.

Continual learning just got a turbo boost: C-Flat Turbo cuts training time by up to 25% without sacrificing accuracy, thanks to a clever gradient-skipping trick.

Achieve real-time (40 FPS at 720p) interactive video generation with minute-long memory consistency using a 5B parameter world model.

iTAG flips the script on causally-annotated text generation, achieving both natural language and high annotation accuracy by treating the causal graph as the target and iteratively refining concept selection.