Haolun Wu 吳昊倫

I am a Ph.D. candidate in Computer Science at McGill University and Mila - Quebec AI Institute. I am delighted in working with Steve Liu, Joelle Pineau, and Laurent Charlin. I also collaborate closely with Fernando Diaz and have him on my supervisory committee. I am currently a visiting scholar working with Sanmi Koyejo at Stanford Trustworthy AI Research (STAIR) Lab. Happy to chat on any related topics.
My research centers on learning from human feedback using ML techniques to make AI systems trustworthy, responsible, and align with human needs. My work on human-AI alignment spans both micro-level aspects (such as personalization and data values) and macro-level aspects (such as social goods and norms). Additionally, I love interdisciplinary research and am particularly interested in applying AI/ML techniques to Education and Psychology. Additionally, I am one of the organizers of the OracleLLM community, where we dive into the frontier of using LLMs as oracles—trusted agents capable of providing reliable, high-level insights. I am also one of the significant contributors to the Test-Time Scaling (TTS) survey, a collaborative effort to understand and push the boundaries of how LLMs can adapt and improve at inference time. I co-founded and co-organize DEFirst, a reading group jointly run at Mila and the Vector Institute, providing an interdisciplinary forum for researchers from academia and industry working on diversity, explainability, and fairness in Information Retrieval, Recommendation, and Trustworthy AI. You can find past talks at our Youtube channel.
During my Ph.D. journey, I interned and collaborated closely with researchers at Microsoft Research, Google Research, and Google DeepMind. My research is generously supported by McGill Graduate Excellence Award, Borealis AI Fellowship, and FRQNT PhD Scholarship (ranked 1st place).
Below is a list of topics covered in part of my projects that I led or highly engaged:
-
▶ 👤 Align AI towards Human Preference
- User modeling and preference elicitation [NeurIPS'24] [ICDE'24] [SIGIR ICTIR'24] [TOIS'23]
- Efficient Personalized LLM Adaptation [ICML'25]
-
▶ 🌍 Align AI towards Social Goods
- Fairness and Inclusivity [SIGIR'22] [TKDE'24] [TOIS'22]
- Knowledge and AI+Education [EMNLP'24] [CHI'24] [AIED'25] [AIED'25]
🚀 New Research Alert: Logits are All We Need to Adapt Closed Models…
— Haolun Wu (@Haolun_Wu0203) March 4, 2025
Very happy my intern work @GoogleAI accepted to #NeurIPS2024 @NeurIPSConf
— Haolun Wu (@Haolun_Wu0203) September 26, 2024
We use Gaussian Process for user modeling to capture multiple interests in an elegant way.
Grateful to my collaborators and mentors @MeshiOfer @masrour_zoghi, @841io, Steve, @Maryam_kar, @CraigBoutilier
My second @MSFTResearch x @Mila_Quebec collaboration...
— Haolun Wu (@Haolun_Wu0203) September 21, 2024
News
🎙️ I am honored to be invited by Prof. Tongliang Liu to give a Rising Star talk on the International Symposium on Trustworthy Foundation Models held at MBZUAI.
Two papers on AI for collaborative learning and decision-making in Education have been accepted to AIED 2025.
One paper on LLM adaptation and alignment has been accepted to ICML 2025.
One paper on Offline model-based optimization (MBO) has been accepted to TMLR 2025.
One paper on LLM personalization with transparency user autonomy has been accepted to The Web Conf 2025 (acceptance rate: 19.8%).
I started my visiting scholar at Stanford University, working with Prof. Sanmi Koyejo on human-centered AI and LLM personalization.
My intern work at Google Research on density-based user representation for multiple interests has been accepted to NeurIPS 2024.
One paper on knowledge extraction using LLMs has been accepted to EMNLP 2024 main conference (Oral, top 7% of all accepted papers).
One survey paper on LLMs for telecommunications has been accepted to IEEE Communications Surveys and Tutorials 2024 (impact factor: 35.6).
One paper on diffusion-based contrastive learning for sequential recommendation has been accepted to CIKM 2024.
One theoretical paper on group-aware search success is accepted to SIGIR ICTIR 2024.
I am ranked in the 1st-place for the prestigious doctoral research scholarship by the Fonds de recherche du Québec – Nature et Technologies (FRQNT) 🎉.
One survey paper on diversity in search and recommendation is accepted to TKDE 2024.
One interdisciplinary collaboration on using linguistic features to reveal learners' psychological patterns accepted to CHI 2024.
One paper on knowledge distillation accepted to ICLR 2024.
I have the honor of being elected as a Lab Representative at Mila, proudly representing the McGill PhD cohort.
I am selected as one of the 10 recipients of the Borealis AI Fellowship, which aims to advance world-leading AI research across Canada, guiding exceptional students and helping them achieve their research goals.
I am happy to be a student researcher at Google Research, New York. I am in Cicero's team and doing research on Mixture-of-Experts (MoE) and knowledge storage/modeling.
Our paper on multi-interest recommendation, Coarse-to-Fine Knowledge-Enhanced Multi-Interest Learning Framework for Multi-Behavior Recommendation, has been accepted by ACM TOIS 2023.
I start my second MSR x Mila research collaboration with the Alexandria team at MSR Cambridge, UK.
I start my research intern at Google Research, Mountain view. I will join Craig Boutitlier's team and do research on multi-interest retrieval in online recommendation systems.
My first (in-person) conference ever! SIGIR 2022 at Madrid 🇪🇸.
I moved to Montreal.