Jiayu Wang

Current video world models may look good on the surface, but they fail to handle critical reasoning tasks, revealing a gap in trustworthiness that could jeopardize robotic manipulation safety.

LLMs can be steered to generate more faithful reasoning chains without sacrificing accuracy using a novel geometric and entropy-based framework, outperforming even GPT-4 in faithfulness detection.

Unified benchmarks reveal the state-of-the-art in simultaneously addressing multiple real-world image degradations like blur, low-light, and rain.

Forget giant models: A carefully trained, quantized SLM can beat proprietary LLMs at aligning with human annotators.

Achieve diffusion-level perceptual quality in monocular depth estimation at 40x the speed, by replacing the slow initial diffusion steps with a fast ViT-based depth map and refining in a compact latent space.

By merging models on the Fisher-Rao manifold, this work achieves stable and accurate LLM merging even with many heterogeneous models, overcoming the representation collapse issues plaguing simpler weight averaging techniques.

SkillOrchestra slashes the learning costs of AI agent orchestration by up to 700x while improving performance by explicitly modeling agent skills and costs, offering a more scalable and interpretable alternative to RL-based methods.