Bowen Yang

Native desktop agents can finally bridge the gap between visual GUI interaction and programmatic code execution by learning a unified action space, achieving state-of-the-art performance on complex real-world tasks.

Robots can now reliably act on human gestures by anchoring interaction evidence to specific object instances, achieving over 2x success on implicit intent tasks compared to large VLM baselines.

Human-robot interaction anticipation models exhibit significant directional bias in cue transferability, with pose corrections boosting performance only when transferring from a less to a more visually diverse dataset.

Intern-S2-Preview-397B not only excels in multimodal scientific reasoning but also enhances biological instruction performance without altering its foundational architecture.

Simulating 8.3 billion diverse personas reveals nuanced user interactions that traditional evaluations miss, transforming how we assess AI systems.

Even the best vision-language models struggle with reliable evaluation of computer-using agents, but OS-Shepherd models offer a low-cost solution that matches their performance.

A unified evaluation framework that simplifies the assessment of LLM-based agents could drastically enhance reproducibility and accelerate research breakthroughs.