Simplifying the latest STEM breakthroughs for all.
Highlights: New framework bridges video generation (ViGen) and motion generation (MoGen) to improve human motion modeling. Introduction of ViMoGen-228K, a large-scale dataset with 228,000 high-quality text-motion and text-video-motion samples. ViMoGen:…
Highlights: Researchers uncover the root cause of instability in RL fine-tuning of large language models. FP16 precision eliminates the mismatch between training and inference phases. The fix is simple, requiring…
Highlights: Researchers introduce the Remote Labor Index (RLI), a comprehensive benchmark for measuring AI automation in remote work settings. The study evaluates AI agents on real-world, economically valuable projects across…
Highlights: Introduces ViMoGen-228K, a dataset of 228,000 diverse human motion samples. Presents ViMoGen, a diffusion transformer integrating video generation insights into motion generation. Develops ViMoGen-light for efficient motion synthesis without…
Highlights: New unified theory integrates Riesz regression, covariate balancing, density-ratio estimation, TMLE, and matching estimators. Introduces Bregman-Riesz Regression for direct debiased machine learning in causal inference. Connects classical econometric methods…
Highlights: Researchers introduce the MME-CoF benchmark for evaluating video-based reasoning capabilities. The popular Veo-3 video model is tested across 12 reasoning dimensions, including geometry and temporal logic. Findings suggest that…
Highlights: Researchers develop an axiomatic characterization of the excess growth rate from portfolio theory. The study reveals deep links between information theory concepts and financial mathematics. Introduces three theorems based…
Highlights: Researchers introduce a novel E-Score framework for evaluating correctness in generative model outputs. The method eliminates vulnerabilities to p-hacking seen in previous p-value-based systems. Provides adaptivity and statistical guarantees…
Highlights: Introduces a meshless framework for solving nonlinear PDE inverse problems on irregular geometries. Utilizes spectral basis parameterization defined on a hyperrectangle enclosing the physical domain. Employs loss-function optimization inspired…
Highlights: New ProMoE framework enhances Mixture-of-Experts (MoE) routing in Diffusion Transformers. Achieves superior performance on ImageNet benchmarks using explicit routing guidance. Introduces two-step routing with conditional and prototypical routing for…