Nan Wang

Robotic world-action models do not need pretrained internet video backbones to scale: training native planning and dynamics modules from scratch across 30,000 hours unlocks strong zero-shot manipulation and drives a 17.7-point gain on an embodiment comprising under 2% of the data.

High binary recognition performance in AMP models fails to predict real-world assay outcomes, revealing critical gaps in current evaluation methods.

Achieving robust tire pattern recognition with limited data, this method leverages dual-branch inference and innovative feature fusion to outperform existing techniques.

A self-evolving critic can reduce confidence estimation errors in LLM agents by up to 54% without requiring any training or ground truth labels.

Jointly estimating the B0 field and distortion-free images directly from k-space data leads to unprecedented improvements in MRI accuracy and detail fidelity, especially in challenging high b-value scenarios.

Scaling up robot data and closing the loop with state decoding and automated reward scoring allows a 2B parameter video world simulator to outperform larger, dedicated robotic world models in real-world policy transfer.

Current multimodal LLMs struggle to understand scientific spectra, but a new benchmark and data processing technique could change that.

Achieve state-of-the-art hyperspectral image fusion by ditching handcrafted assumptions and protocols for a generalizable flow-matching framework.

Finally, a real-time 4D world simulator exists that allows for consistent and controllable scene evolution from a single monocular video.

Unlock hidden app functionality: EpiDroid's dependency-aware recomposition boosts code coverage by 3-4x compared to standard exploration, even with the same budget.