Jing Wu

Today's best multimodal LLMs are surprisingly inept at using tools to solve agricultural tasks, struggling with everything from planning to error recovery.

Ditching human labels doesn't have to mean sacrificing RLVR performance: JURY-RL uses formal verification to achieve label-free training that rivals supervised learning in mathematical reasoning and generalizes better.

Latent reasoning can beat explicit Chain-of-Thought – but only if you force it to learn causal dynamics via a visual world model, not just language.

Camera array super-resolution gets a boost: a new self-supervised method leverages both multi-image-to-single-image and multi-image-to-multi-image techniques to generate sharper, more detailed images.

A diffusion model can generate high-quality synthetic chromosome images, boosting anomaly detection by nearly 14% F1 score and reducing reliance on scarce real-world abnormal samples.

Injecting physics-based priors derived from MLLMs at decoding time significantly boosts weather forecasting accuracy and stability, even in long autoregressive rollouts.

OpenPangu-7B inference on NPUs gets a serious speed boost via a custom-tailored speculative decoding scheme.