Xin Yu

Ph.D. Candidate in Statistics

The Pennsylvania State University

I am currently a Ph.D. candidate in Statistics at The Pennsylvania State University and am honored to be advised by Prof. Lingzhou Xue. I earned my master's degree from Penn State in 2025 and my bachelor's degree in Statistics from Nankai University in 2023. I was also fortunate to complete my undergraduate thesis under the supervision of Prof. Changliang Zou.

I study how AI systems learn from feedback and experience, with a focus on self-improving agents and multimodal understanding.

Research Focus

Research directions & related papers
An agent learns from a failed check An illustrative loop: the agent tries a task, receives a failed check, stores a lesson in memory, and uses that lesson on the next attempt. TRYREFLECTIMPROVE TASK Agent Checks Memory retain the lesson Try an approach. Turn a failed check into a lesson. Use that lesson on the next attempt. 01 / SELF-IMPROVING AGENTS

Recursive Self-Improvement

Autonomous agents that evaluate their own work, retain useful experience, and improve their workflows across tasks.

From a product page to structured attributes Illustrative product extraction: a model connects an image of a blue shirt with the text Blue linen shirt and $39, then produces item, color, and price fields. OBSERVECONNECTEXTRACT Blue linen shirt $39 vision + language Attributes itemcolorprice shirtblue$39 Find evidence in images and text. Connect the evidence across modalities. Extract grounded, structured information. 02 / VISION & LANGUAGE

Multimodal Understanding

Models that connect visual and textual evidence to extract structured information and reason about real-world content.

News

Earlier updates 5 more
  • I started my internship at TikTok, focusing on multimodal understanding and exploring continual learning and agent-based automated error correction. ✨
  • I presented a poster at NeurIPS 2025 at the San Diego Convention Center. ☀️
  • I presented a poster at ICML 2025 at the Vancouver Convention Center. ☀️
  • I joined ByteDance Seed Infra in Seattle for a summer internship, working with Cong Xie and Zhi Zhang. 🧑‍💻
  • I’m thrilled to share that I’ve completed my five-month internship at Microsoft Research Asia (MSRA). I’m grateful to have worked with Yan Lu and Xiaoyi Zhang and honored to have received the Stars of Tomorrow award! 🧑‍💻

Selected Publications

See Google Scholar for an updated list of publications.

Industry Experience

Google

Aug. 2026 – Nov. 2026

Student Researcher · Sunnyvale, CA

  • Coding Agents
  • Agent Post-Training
  • Recursive Self-Improvement

Meta · Feed Discovery

May 2026 – Aug. 2026

Research Scientist Intern · Menlo Park, CA

  • Generative Recommendation
  • Tool-Augmented Agents
  • Agentic RL
  • GRPO

TikTok · Ads Integrity

Nov. 2025 – May 2026

Research Scientist Intern · San Jose, CA

  • Multimodal Understanding
  • Ads Integrity
  • Self-Rewarding
  • On-Policy Distillation

ByteDance · Seed

May 2025 – Dec. 2025

Research Scientist Intern · Seattle, WA

  • Efficient RL Training
  • VERL / FSDP
  • GRPO
  • Diffusion Preference Optimization

Microsoft Research Asia

Jan. 2023 – May 2023

Intelligent Multimedia Group Intern · Beijing, China

  • Efficient Vision Models
  • Imbalanced Learning
  • OCR

Education & Recognition

Education

  • The Pennsylvania State UniversityPh.D. candidate in Statistics · CurrentMaster’s degree · 2025
  • Nankai UniversityBachelor’s degree in Statistics · 2023

Selected recognition

  • MLSS NYC travel awardFull travel award · Columbia University · 2026
  • Stars of TomorrowMicrosoft Research Asia · 2023

Service

Reviewer

Conferences: ICLR · ICML · NeurIPS · AAAI · ECCV · KDD · ICASSP · ACL ARR

Journals: TNNLS · TCSVT · TOSN · Pattern Recognition · IJDSA

Full reviewing record
  • The Fifteenth International Conference on Learning Representations (ICLR 2027): 5.
  • ACL Rolling Review (ARR), October 2026: 2.
  • The Fourteenth International Conference on Learning Representations (ICLR 2026): 5.
  • 2026 IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP 2026): 2.
  • 2025 IEEE Transactions on Circuits and Systems for Video Technology: 1.
  • 2026 Pattern Recognition: 2 manuscripts (3 review rounds).
  • 2026 IEEE Transactions on Neural Networks and Learning Systems: 3.
  • International Journal of Data Science and Analytics: 1.
  • KDD 2026 AI4Sciences Track: 3.
  • Neural Information Processing Systems (NeurIPS 2026): 4.
  • NeurIPS 2026 Workshop on Meta-Agents: 2.
  • 2026 ACM Transactions on Sensor Networks: 1.
  • AAAI 2027: 6.
  • Forty-Third International Conference on Machine Learning (ICML 2026): 6.
  • ECCV 2026: 2.

Counts refer to distinct manuscripts reviewed; revision rounds are noted separately.

Current Reviewing Service
  • KDD 2027 AI4Sciences Track Cycle 1: Reviewer (ongoing).
Conference / Seminar Organizer
Teaching Assistant / Consultant
GET IN TOUCH

Let’s talk research.

Statistical machine learning, foundation models, and learning agents.

Visitors

Loading visitor map…