Hao Guo

SkillChain transforms the way image-based AI assistants adapt to user intents, achieving substantial gains in response quality and user retention.

Transforming edge posteriors from diffuse uncertainty into structured clarity, Diff-prior revolutionizes how we infer interaction graphs in complex systems.

Subterranean agents—LLMs with agentic workflows compiled directly into their weights—can match the performance of state-of-the-art orchestration frameworks while slashing costs by two orders of magnitude.

Agent orchestration frameworks might be overkill: simply including the entire procedure in the system prompt yields better performance on procedural tasks.

Video LLMs can ace individual traffic video questions but still fail spectacularly at subtle counterfactual reasoning, revealing a critical blind spot for safety-critical applications.