Yuxuan Li 李宇轩
Humans are increasingly letting AI make decisions on their behalf. My research asks how we can , , and AI agents for consequential decision making, whether they directly act as decision makers or simulate behavior to inform human decisions.
EvaluateI build benchmarks and evaluation methods to test AI agents in consequential decision-making.
- HiddenBenchICML 2026
- Actions Speak Louder than WordsFAccT 2025
- Spontaneous Giving, Calculated GreedEMNLP 2025
- SP-ABCBenchNAACL 2027 · under review
- CAVEATICLR 2027 · under review
- Physically Grounded Agent SimulationOngoing
ImproveI train LLMs and develop agent architectures and reasoning processes to make consequential decisions more robust.
- CAVEAT-Harness + CAVEAT-27BICLR 2027 · under review
- Self-Evolving Agents TeamsOngoing
- S&P Theory-Informed Agent SimulationNAACL 2027 · under review
IntegrateI design systems, interfaces, and decision processes for integrating AI agents into human and institutional decision-making.
- WhatIf: Interactive Large-Scale Agent SimulationUIST 2026
- LLM Agent Simulations for PolicyPoliSim @ CHI 2026
- Recursive Simulation ModelingCHI 2027 · under review
- Scalable Human Oversight of Agent CollectivesOngoing
I am a third-year PhD student in the Human-Computer Interaction Institute (HCII) at Carnegie Mellon University's School of Computer Science, advised by Hirokazu Shirado and Sauvik Das. I received my BS in Computer Science from Tsinghua University in 2024. I have worked as a research intern at Microsoft Research AI Frontiers, Microsoft AI, and UC Berkeley. My first-author work has been published at leading venues including ICML, EMNLP, UIST, and FAccT, and received . I have given invited talks at institutions worldwide, including CMU, Georgia Tech, Tsinghua, KAIST, HKU, and Microsoft.
- Smarter AI Models Show Selfish Behavior in StudyThe Chosun Ilbo
- AI showing signs of selfishness, researchers warn of troubling trendThe News International
- When AI LLMs think more, groups suffer, CMU study findsGeneva Internet Platform
- Yapay zeka akıl yürütme kapasitesi arttıkça bencilleşiyorNTV
- Is AI Becoming Selfish?Carnegie Mellon University
- CyLab researchers to present at ACM CHI 2026CyLab, Carnegie Mellon University
- Your friendly neighbourhood AI just got smart and surprisingly selfishDigital Trends
- CMU researchers find 'selfish' traits in cutting-edge AITribLIVE
- Your friendly neighbourhood AI just got smart and surprisingly selfishYahoo News
- Smarter AI models show more selfish behaviorPsyPost
- The smarter AI gets, the more selfish it becomes, study warnsKnowridge
- Advanced AI Is Becoming More Selfish As It Learns To ReasonDevDiscourse
- Carnegie Mellon Study Finds Advanced AI Systems Tend to Prioritize Self-InterestGeneOnline
- Carnegie Mellon Study Finds Advanced AI Becomes More Self-InterestedDigital Information World
- Smarter AI, Selfish AI: Carnegie Mellon Study Reveals Downside Of Machine ReasoningAZoAI
- AI may turn self-centered as it becomes more intelligentAsianet Newsable
- Artificial Intelligence is becoming increasingly selfishEuro Security
- AI 隊友不如人意,研究揭深層推理大模型「只顧自己」TechNews
- AI becomes more selfish as reasoning capacity increasescnBeta
- Levante-EMV
- ABC Gazetesi
- Arsa Technology
- Mundiario
- Revista Perspectiva
- Referans Türk
- Bioengineer.org
- Scienmag
- InnoLabs
- Xenospectrum
- Kosmicki Glasnik
- Rexmolon
- Tunceli Emek
- Cong Dong AI
- SciTke
- JDYX
- INF News
- INF News
- Haber Cephesi
- The Daily Horizon
- Wisdom Superman
- E-Gündem
- Kadın Elit
- AI Buzz
- Hi-Network
- AI Search
- DeepTechBytes
- The420