How to Tell If Your AI Truly Generalizes
Humanity developed the best AI model on Earth and brought it to Mars. How do we make sure that it still works when there's no one around to fix it when it breaks?
blank
Peng Qi is an AI researcher working on natural language processing, machine learning, and multimodal agents.
Latest posts
Why You Should Stop Using HotpotQA for AI Agents Evaluation in 2025
We published HotpotQA, a groundbreaking multi-step question answering dataset in 2018, which has since motivated and facilitated numerous AI agent research works. But you should probably reconsider using it for your AI agents research in 2025.
AI is the New Rocket Science
AI science of today has astonishing similarities to rocket science in its prime days, if one pays close attention to history. What are some of these, and what can the history of rocket science tell us about the fate of AI?
What do industry researchers do, anyway? Part 2 – What do they do when they are not publishing
Many people seem to "disappear" from the academic publication scene once they graduate from their PhD and join industry research labs. What are they actually doing when they are not publishing? What do they have to show for their research work if they don't publish?
What do industry researchers do, anyway? Part 1 – How to choose a team
You are about to embark on a journey as an industry researcher, or are ready for your next adventure. What can help you choose the most suitable team for yourself?
What do industry researchers do, anyway? Part 0 – Academia vs Industry
What does an industry researcher actually do, and how does it differ from academia? Should I go for academia or the industry after my PhD?
What does an area chair actually do, anyway?
What does serving as an AC actually entail, and what can be improved in our review process?
Teaching Conversational NLP Systems to Ask Informative and Specific Questions
Asking good inquisitive questions in a conversation requires inferring about others' knowledge. Can we train NLP systems to do this?
Answering Complex Open-domain Questions at Scale
The NLP community has made great progress on open-domain QA, but our systems still struggle to answer complex open-domain questions in an large collection of text. We present an efficient and explainable method for enabling multi-step reasoning in these systems.
Pinyin Cheatsheet for (Mostly American) English Speakers
Cheatsheet to pronounce pinyin (especially names romanized in pinyin) for English speakers (mostly applicable to the American accent).