Yi Zheng

The paper systematically measures the risk of current image-to-3D models generating harmful geometries, finding that these models are effective at reconstruction and existing safeguards are insufficient.

The paper introduces AgenticVBench, a comprehensive benchmark of 100 real-world video post-production tasks, and finds that even the best AI agents perform significantly worse than human experts on these complex, multi-modal tasks.

The paper proposes EKSFT, a selective fine-tuning method that masks high-entropy or high-KL divergence tokens during Supervised Fine-Tuning (SFT) to prevent distribution shift and improve subsequent Reinforcement Learning (RL) performance.

InfoAtlas is a foundation model that estimates statistical mutual information (MI) in a single forward pass, achieving state-of-the-art accuracy with a massive speedup compared to traditional iterative neural estimators.