Hua Wei

@KDD2026 is introducing an AI for Sciences track (Abstract deadline: Feb 1) and this feels bigger than “one more track.” kdd2026.kdd.org/ai4sciences-tr… A few interesting parts: 🧩 It explicitly welcomes 2-page extended abstracts of already-published journal work ❌Overly complicated frameworks for the sake of technical innovation are discouraged. That feels like KDD rediscovering its roots👉new scientific problems > algorithmic novelty.

New paper alert (AAAI’26) For self-driving cars, trajectory prediction evaluation has long relied on metrics like ADE/FDE, which often say little about how errors actually affect downstream planning and driving behavior. Together with @Honda Research Institute, we take a step toward evaluation that reflects real-world impact, not just prediction accuracy. Paper：arxiv.org/pdf/2512.12211 Code：github.com/honda-research… Please stop by our poster and talk to @manishsaroya if you are traveling to Singapore 🇸🇬

🚀 We are hiring Research Interns ! We are looking for Master's and PhD students (final year) with a strong background in AI / ML / NLP who want to work on cutting-edge AI systems alongside Mistral AI researchers. 📍 Location: Paris, London, or Palo Alto ⏳ Duration: 4-6 months 🎓 For Master's students: opportunity to continue with a PhD at Mistral after the internship (via the CIFRE program). Links below:

🌍 If you mix languages in your query, you are likely using a degraded AI agent ⚠️Because tool execution quietly assumes English ⚙️🇺🇸. Our recent analysis shows that multilingual degradation in tool-calling agents largely comes from language-invariant execution assumptions. We introduce a diagnostic benchmark (including Chinese 🇨🇳, Hindi 🇮🇳, and Igbo 🇳🇬), and show why inference-time “fixes” only go so far. TL;DR: If AI agents are deployed globally 🌐, execution conventions must be designed globally too. Paper: arxiv.org/pdf/2601.05366 #AgenticAI #LLMs #MultilingualAI #LLMAgents #ToolCalling #NLP #AISystems

🚀 Our open-sourced Instructional Agents - an AI-powered instructor assistant - has been accepted to the #EACL2026 Main Conference! After talking with many educators, we kept hearing the same question:👉 Can AI help instructors keep courses current, pedagogically coherent, and low-effort? Turns out: yes. With Instructional Agents, building a new course can cost < $1, while supporting end-to-end course creation (including syllabus, slides, talking notes, assessments, etc.) without sacrificing instructional quality. Project: hyan-yao.github.io/instructional_… Paper: arxiv.org/abs/2508.19611 gde: hg 🔑 Key Features of Instructional Agents: 🤖 Collabotated Role-Based Multi-Agents Instead of a single model, we simulate a full instructional team. Specialized agents take on the roles of Teaching Faculty, Instructional Designer, Course Coordinator, Program Chair, and Teaching Assistant, enabling a clear division of labor. 🧠 Grounded in Instructional Framework The entire workflow is grounded in the ADDIE instructional design framework (Analyze, Design, Develop), ensuring alignment across learning objectives, syllabi, and assessments. 💬 Human in Control Automation does not mean losing control. The system supports four modes: Autonomous, Catalog-Guided, Feedback-Guided, and Full Co-Pilot, so instructors can choose the level of oversight they want. 📄 Full package Generation From a topic prompt to classroom-ready assets: comprehensive syllabi, LaTeX-based slides, scripts, and rigorous assessments- all in one unified pipeline.

@EugeneVinitsky Algorithmic martyrdom for students 🫡

@realhuawei I am not. I show up here every couple of days to post bad vibes so that when my students post papers, my reposts do not get dinged

Super excited to announce my new paper on this website! Thank goodness I'm not the sort of person that trolls would use Grok to make non-consensual explicit imagery of, that'd feel super bad

@prof_yz @LongchaoDa @ArianeMiddel 🔥

@realhuawei @LongchaoDa @ArianeMiddel #dreamitpossible ❤️

Heat wave survival tip: Walk like a local, route like an AI. 🌞➡️🌳 Proud of my amazing students like @LongchaoDa and collaborators @prof_yz @ArianeMiddel for making this happen

🌳 Shade can lower surface temps by 20°F — sometimes enough to protect health. At @ASU, Hua Wei’s AI tools map and predict shade, helping people find cooler routes and guiding cities to design greener spaces. #ASUEngineering @realhuawei news.asu.edu/20250925-scien…

For everyone who ever asked me "how big is your dept faculty?"--well just this big👇.. @SCAI_ASU

🚨 Big news! Our paper "Fully Heteroscedastic Count Regression with Deep Double Poisson Networks (DDPN)" was just accepted to #ICML2025! 🥳🎉 Let me tell you why we're excited — especially if you care about trustworthy AI and count data. 👇🧵

@VikashGayah @SCAI_ASU Thanks Vikash!

@realhuawei @SCAI_ASU Congrats!!!

Honored to receive the NSF CAREER Award for my project on Sim-to-real Transfer for Decision Making! A HUGE thanks to my students, collaborators, mentors, panelists, and @SCAI_ASU for their support!

🎉HUGE congrats to Longchao Da, a doctoral student working under the supervision of @realhuawei, who received the 🏆SDM 2025 Doctoral Forum Best Poster Award at the @TheSIAMNews International Conference on Data Mining held last week in Alexandria.

@yuz9yuz @SIAMDataMining @dmguiuc @BowenJin13 Nice tutorial!

Had a great time giving a tutorial at #SDM2025 @SIAMDataMining with Prof. Jiawei Han @dmguiuc and @BowenJin13!

Looking forward to meeting old/new friends there!

A team led by @PauloShakAS, an associate professor in @SCAI_ASU, has received a $1.7 million research grant from @AZRegents to develop software powered by #AI that will assist emergency responders in predicting and managing disaster scenarios. fullcircle.asu.edu/research/ai-fo…