Bio

• Can Xu is currently a Senior Research Scientist at Microsoft AI. He has contributed core techniques for Microsoft XiaoIce, Bing, WizardLM and Phi-3.
• Can Xu led the Research & Development of WizardLM family models: WizardLM 1&2, WizardCoder, WizardMath
• Can Xu has published 40+ papers in top-tier international conferences, such as NeurIPS, ACL, ICLR, EMNLP, AAAI etc. He has received over 3300 citations on Google Scholar.
• His research interests include large language models, dialogue systems, and information retrieval.
• He received bachelor’s degree in computer science from Sichuan University and master’s degree (advisor was Jingsong Yu) in computer technology from Peking University.

Recent News

• 3 long papers accepted by EMNLP 2024 main conference!
• [2024.04] Release WizardLM-2, which outperforms GPT-4 on MT-Bench, GPT4-Turbo on AlpacaEval 2.0 and Claude 3 Sonnet on Arena-Hard.
• [2024.01] Our papers WizardLM and WizardCoder got accepted to ICLR 2024.
• [2023.08] WizardCoder achieves the 1st-rank on Multilingual Code Models Evaluation Leaderboard. [澎湃报道]
• [2023.06] WizardLM achieves the 1st-rank of the opensource models on Standford AlpacaEval Leaderboard.
• [2023.04] Release WizardLM expertized in following complex instructions. [Github] (Over 9K Stars) [WizardLM Pages] [Huggingface] [Beeboom: 12 Best LLMs in 2024] [华尔街见闻]
• 1 long papers accepted by ACL 2023!
• 1 long papers accepted by CVPR 2023!
• 3 long papers accepted by ICLR 2023!
• 1 paper about Peer-Contrastive Learning for Unsupervised Sentence Embeddings accepted by EMNLP 2022!
• 1 paper about Fine-Grained Contrastive Learning for Dialogue Response Selection accepted by INTERSPEECH 2022!
• 1 paper about Stylized Knowledge-Grounded Dialogue Generation accepted by NAACL 2022!
• 4 long papers(3 Accept, 1 Accept Findings) accepted by ACL 2022!
• 1 paper about Unsupervised Cross-Domain Adaptation for resposne selection accepted by WSDM 2022!
• Received a NeurIPS 2021 Outstanding Reviewer Award
• 1 paper about Fine-grained Controllable Text Generation accepted by NeurIPS 2021!
• 2 papers(1 Accept, 1 Accept Findings) accepted by EMNLP 2021!
• 3 papers(2 Accept, 1 Accept Findings) accepted by ACL 2021!
• 1 paper about Open Domain Dialogue Generation with Latent Images accepted by AAAI 2021!
• 1 paper about zero-resource knowledge-grounded dialogue generation accepted by NeurIPS 2020!
• 3 papers(2 Accept, 1 Accept Findings) accepted by EMNLP 2020!
• 1 paper about low-resource knowledge-grounded dialogue generation accepted by ICLR 2020! In this work, we present how to achieve sota performance on Wizard of Wikipedia with only 1/8 training examples.
• 2 papers accepted by EMNLP 2019!

Hiring

• looking for highly self-motivated students to work with me as research interns

Working Experiences

• June. 2024 - Now, Senior Research Scientist, Microsoft AI.
• July. 2020 - Now, Research Scientist, Natural Language Processing Group, Microsoft Software Technology Center at Asia.
• July. 2020 - Apri. 2018, Research Scientist, XiaoIce Research Group, Microsoft Software Technology Center at Asia.
• Apri. 2017 - Apri. 2018, Associate Researcher, Natural Language Computing Group, Microsoft Research Asia.

Academic Services

Program Committee for

• NeurIPS 2021,2022
• ACL 2020,2021,2022
• SIGIR 2021,2022
• AAAI 2021,2022
• EMNLP 2020,2021
• ICLR 2021,2022,2023

Senior Program Committee for

• IJCAI 2021