Zeyu Li

Even state-of-the-art vision-language models frequently lie and hallucinate when playing social deduction games, raising serious questions about their reliability in real-world applications requiring grounded reasoning.

Today's best AI agents can only solve 55% of real-world academic tasks that university students find challenging, revealing a significant gap between current AI capabilities and the demands of academic workflows.

LLMs can leapfrog state-of-the-art scientific algorithms and human-designed solutions, but only if you scale the evaluation loop, not just the model.

CGRA performance jumps by 2.7x thanks to NEURA, a compilation framework that elegantly transforms control flow into dataflow.

LLMs struggle with conflicting medical evidence, but a clever two-stage agentic approach can reconcile discordant signals while preserving patient privacy.

LLMs can't even reproduce published physics papers end-to-end, with the best model scoring only 34% on a new benchmark designed for this purpose.