Yuyao Wang

Yuyao Wang

Ph.D. Candidate in Statistics @ Boston University

Howdy! I’m a Statistics Ph.D. candidate at Boston University, advised by Debarghya Mukherjee and Luis Carvalho. Before BU: M.A. Statistics at Columbia, B.S. Mathematics at Shandong University, with a year at AMSS, Chinese Academy of Sciences.

I work on transfer learning and representation learning — optimal transport, graph methods, multimodal models — with a focus on what holds up when data is scarce, high-dimensional, and non-IID. The question I keep coming back to: how do you reuse what a model already knows when the world won’t sit still? Part of the answer is knowing when transfer provably works — minimax rates, oracle inequalities, safe-transfer criteria. The other part is the cases where structure itself is the obstacle: aligning graphs and manifolds without known correspondence, warm-starting policies in environments that keep moving, and specializing pretrained LLMs and VLMs without letting them overfit or drift out of alignment.

For anyone who wants a gentler entry point, I’ve put together beginner-friendly slide decks on my main directions: transfer learning · graph learning · optimal transport · LLMs for time series. I’ve been lucky to learn from Zhanxing Zhu (STGCN) and Yongshun Gong, whose work on spatio-temporal structure shaped how I think about heterogeneous, evolving data. Outside of research, I’m drawn to building things with atmosphere — cinema, memory, digital experiences that feel like something rather than just work ✨

🔥 News

  • 2026.05: 🎉 My co-first-author paper “From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space” was selected for an Oral presentation (top 0.5%) at (ICML 2026)!
  • 2026.05: 🎉 My co-authored paper “Network Perturbation Aggregation for Graphon Estimation” has been accepted by SLADS!
  • 2026.04: 🎉 My co-first-author paper “From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space” is accepted by (ICML 2026) and selected as a Spotlight!
  • 2026.04: 🚀 I’ll be joining Amazon as an Applied Scientist this summer, based in the Bay Area, California!
  • 2026.04: 🎉 I am honored to receive the Dean’s Dissertation Fellowship from the Graduate School of Arts and Sciences!
  • 2025.09: 🎉 My first-author paper “Transfer Learning on Edge Connecting Probability Estimation Under Graphon Model” is accepted by (NeurIPS 2025)!
  • 2025.08: 🎉 My co-authored paper “Cross-Domain Hyperspectral Image Classification via Mamba-CNN and Knowledge Distillation” is accepted by (IEEE TGRS 2025)!

📝 Publications

Leading Author

GTrans NeurIPS 2025 Transfer Learning on Edge Connecting Probability Estimation Under Graphon Model  Paper Poster Slides Code
  • First graphon-level transfer without node correspondence — aligns graphs via Gromov–Wasserstein and transfers edge structure nonparametrically.
  • Residual smoothing unlocks small/sparse targets with convergence & stability guarantees; SOTA on link prediction and graph classification.
TESS ICML 2026 Oral From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space  Paper Slides Code
  • Bridges the text–time-series modality gap by translating free-form text into interpretable temporal primitives—distribution shift, volatility, shape, and lag—instead of directly fusing noisy token embeddings.
  • Confidence-aware semantic conditioning injects reliable primitives into a Transformer forecaster, achieving robust gains under event-driven non-stationarity and up to 29% error reduction.
Phase Transition Under Review Phase Transition in Nonparametric Minimax Rates for Covariate Shifts on Approximate Manifolds  arXiv Poster Slides Code
  • New minimax theory for "near-manifold" shift: exposes a sharp phase transition controlled by the support gap between target and source neighborhoods — unifying multiple geometric-transfer regimes.
  • Ratio-free, adaptive estimator: achieves near-optimal, dimension-adaptive rates without density ratios and without assuming known geometry (works under approximate manifold mismatch).
SCOT Under Review SCOT: Multi-Source Cross-City Transfer with Optimal-Transport Soft-Correspondence Objectives  arXiv Slides
  • Sinkhorn entropic-OT coupling enables many-to-many region alignment across cities — no node matching required.
  • OT-weighted contrastive loss + target-aware prototype hub prevents collapse and scales cleanly to multi-source heterogeneity.

Co-author

Net-Paging SLADS 2026 Network Perturbation Aggregation for Graphon Estimation  Paper Code
  • Introduces Net-Paging, a perturbation–aggregation framework that generates multiple graphon-preserving networks from a single observed graph to reduce estimation variance.
  • Provides a closed-form bias correction and theoretical guarantees showing that aggregation improves MSE while preserving the convergence behavior of the base estimator.
MKDNet IEEE TGRS 2025 Cross-Domain Hyperspectral Image Classification via Mamba-CNN and Knowledge Distillation  IEEE Slides
  • Hybrid spectral–spatial modeling for domain shift: integrates a Mamba-based global spectral encoder with CNN local feature extraction, capturing long-range dependencies while preserving fine-grained spatial structure.
  • Dual-level transfer via distillation + graph alignment: performs teacher–student knowledge distillation for distribution alignment and OT-guided graph consistency across domains, yielding robust cross-domain generalization under severe spectral mismatch.
SSGP Under Review Semantic Scientific Graph Pruning for Reliable Agentic Paper Reproduction  arXiv
  • SSGP prunes dense scientific graphs into task-adaptive subgraphs via rank-based ensemble scoring — drastically shrinks agent search space.
  • Reuse–patch execution + confidence-weighted aggregation boosts reproducibility, stability, and success rate of LLM scientific agents.

🤖 LLM & DS Projects

LLM Alignment & Evaluation
AlignDPO RAGAudit
DPO · IPO · KTO · QLoRA — NLI · SelfCheckGPT · semantic entropy 		Retrieval & Inference
GraphRAG Adaptive RAG DraftVerify HQQ
entity graph + CLIP · query routing — speculative decoding · 1-bit quantization
Causal Inference & Experimentation
CausalLens Congestion Pricing A/B Testing
DoWhy · Double ML · Causal Forest — CS-DiD · Synthetic DiD · 12M+ NYC TLC 		NLP
Financial Sentiment Spam Detection
DistilBERT fine-tuning · TF-IDF · Naive Bayes
Computer Vision
Dog Classification Mask Detection
VGG16 · ResNet50 · transfer learning · Grad-CAM 		Statistical Modeling
Bayesian Logistic Time Series Credit Risk Segmentation Recommender Airbnb Dashboard
RStan · Spike-and-Slab MCMC · SARIMA · XGBoost · R Shiny

📖 Educations

Boston University  ·  Ph.D. in Statistics  ·  2021.09 – present
Columbia University  ·  M.A. in Statistics, Data Science Track  ·  2019.09 – 2020.05
Shandong University  ·  B.S. in Mathematics  ·  2015.09 – 2019.06
Chinese Academy of Sciences  ·  Jointly Supervised Talent Program, AMSS  ·  2018.05 – 2019.06

💻 Internships

Plymouth Rock

Data Scientist Intern · Plymouth Rock Insurance
📍 Boston, MA  ·  🗓️ May 2025 – Aug 2025

  • Architected an end-to-end AWS SageMaker pipeline for property-level loss prediction using an XGBoost Tweedie model on multi-million-policy data, lifting Gini by +4.3% over the production baseline and directly improving underwriting risk segmentation.

  • Pioneered an LLM-powered visual risk scoring system combining GPT-4o multimodal reasoning with Google Street View imagery to capture previously unobservable property features (roof condition, surroundings, hazards); integrated outputs into downstream actuarial pricing models as a novel signal layer.

  • 📎 For a high-level, non-confidential summary of this work, see the Home Insurance slides.

✨ My Apps

A quiet collection of cinematic, atmospheric, and emotionally resonant side projects — part digital keepsakes, part memory-keepers.  See all →

Wilderness
🌲 Wilderness 		MBTI Vibe
✨ MBTI Vibe 		What If Cinema
🎬 What If Cinema 		Letters from the Screen
✉️ Letters from Screen 		If You Disappeared
✈️ If You Disappeared
Souvenirs
🎟️ Souvenirs 		The Map of Me
🗺️ Map of Me 		A Room in Macondo
🦋 A Room in Macondo 		Say It Like a Classic
✒️ Say It Like a Classic 		The Boston Archive
🏛️ Boston Archive

🎖 Honors

Boston University  ·  Dean’s Dissertation Fellowship (2026) · Ralph B. D’Agostino Fellowship (2025) · Outstanding Teaching Award (2025) Shandong University  ·  Outstanding Graduate (2019) · First-Class Scholarship (2018) · Outstanding Student Leader (2018)
National  ·  Hua Loo-Keng Scholarship (2018) · National Gold Award, Internet+ Innovation & Entrepreneurship Competition (2018)

📝 Service & Teaching

Presentations  ·  CIKM 2024, NeurIPS 2025, ICML 2026
Reviewer  ·  CIKM 2025, ICME 2026, ICML 2026, KDD 2026, KDD 2027
Instructor @ Boston University  ·  MA 582 Mathematical Statistics, MA 113 Elementary Statistics
TA @ Boston University  ·  MA 575 Generalized Linear Models, MA 582, MA 415 Data Science in R, MA 214 Applied Stats

🎨 Interests

🎵 Mandarin R&B loyalist — Leehom Wang, David Tao, Khalil Fong🦋, Dean Ting

🎹 Trained in piano, calligraphy, and ink painting

🏞️ National park lover · 🫧 lake admirer · 🌅 opacarophile — welcome to my Gallery