Howdy! I’m a final-year Statistics Ph.D. candidate at Boston University, advised by Debarghya Mukherjee and Luis Carvalho. Before BU, I studied statistics at Columbia (M.A.) and mathematics at Shandong University & Chinese Academy of Sciences (B.S.), and worked with Zhanxing Zhu and Yongshun Gong on spatio-temporal data. Currently, I work on transfer and representation learning for modern AI systems, from multimodal foundation models to LLM agents. The question I keep coming back to is how you reuse what a model already knows when the world won’t sit still:

• LLM agents. Post-training agents for changing environments: RL from sparse rewards, on-policy self-distillation, and spatio-temporal reasoning.
• When transfer provably works. Minimax rates, oracle inequalities, and safe-transfer criteria.
• When structure is the obstacle. Optimal transport, graph, and multimodal methods for aligning heterogeneous data.

A gentler entry point, in slides: LLM agents · transfer learning · graph learning · optimal transport · LLMs for time series

🔥 News

• 2026.05 — TESS (co-first) got an Oral at ICML 2026, top 0.5% of submissions!
• 2026.05 — “Network Perturbation Aggregation for Graphon Estimation” (co-author) is in at SLADS.
• 2026.04 — Heading to Amazon as an Applied Scientist this summer, Bay Area bound.
• 2026.04 — Honored to receive the Dean’s Dissertation Fellowship from BU’s Graduate School of Arts and Sciences.
• 2025.09 — GTrans (first author) accepted at NeurIPS 2025.
• 2025.08 — “Cross-Domain Hyperspectral Image Classification” (co-author) accepted at IEEE TGRS.

📝 Publications

First author

Transfer Learning on Edge Connecting Probability Estimation Under Graphon Model

Graphon-level transfer without node correspondence: aligns graphs via Gromov–Wasserstein, then transfers edge structure nonparametrically, with residual smoothing that unlocks small and sparse targets.

NeurIPS 2025 / Paper / Poster / Slides / Code

From Text to Forecasts: Bridging Modality Gap with Temporal Evolution Semantic Space

Translates free-form text into interpretable temporal primitives (distribution shift, volatility, shape, lag) and conditions a Transformer forecaster on them, reducing error by up to 29% under event-driven non-stationarity.

ICML 2026 (Oral) / Paper / Slides / Code

Phase Transition in Nonparametric Minimax Rates for Covariate Shifts on Approximate Manifolds

Minimax theory for near-manifold covariate shift, exposing a sharp phase transition governed by the support gap, together with a ratio-free estimator that adapts to unknown intrinsic dimension.

Under review / arXiv / Poster / Slides / Code

SCOT: Multi-Source Cross-City Transfer with Optimal-Transport Soft-Correspondence Objectives

Sinkhorn entropic-OT coupling gives many-to-many region alignment across cities with no node matching, paired with an OT-weighted contrastive objective that resists collapse under multi-source heterogeneity.

Under review / arXiv / Slides

Co-author

Network Perturbation Aggregation for Graphon Estimation

Generates graphon-preserving networks from a single observed graph and aggregates them to cut estimation variance, with a closed-form bias correction and guarantees that the base estimator's convergence is preserved.

SLADS 2026 / Paper / Code

Cross-Domain Hyperspectral Image Classification via Mamba-CNN and Knowledge Distillation

Pairs a Mamba global spectral encoder with CNN local features, then transfers across domains through teacher–student distillation and OT-guided graph consistency under severe spectral mismatch.

IEEE TGRS 2025 / Paper / Slides

Multi-Scale Based Cross-Modal Semantic Alignment Network for Radiology Report Generation

Aligns radiology images and reports across scales via sparse local attention at multiple dilation rates, coupled with a gated shared-memory alignment module and a multi-label contrastive objective.

IEEE SMC 2025 / Paper

What Papers Don't Tell You: Recovering Tacit Knowledge for Automated Paper Reproduction

Prunes dense scientific graphs into task-adaptive subgraphs by rank-based ensemble scoring, shrinking the search space for LLM agents and raising the success rate of automated paper reproduction.

Under review / arXiv

🤖 LLM & DS Projects

📖 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

Applied Scientist Intern · Amazon · Summer 2026
_{LangChain agent with LLM-based heuristic learning that turns request-level attribution into ranked traffic-blocking policies; MIMO forecasting on large-scale HTTP logs, benchmarking tabular foundation models against Chronos-2.}

Data Scientist Intern · Plymouth Rock Insurance · Summer 2025
_{Multimodal property risk scoring with GPT-4o and Street View imagery; XGBoost Tweedie loss model on SageMaker (+4.3% Gini). Slides}

🎖 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 and 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  ·  Mathematical Statistics (MA 582), Elementary Statistics (MA 113)
Teaching Fellow  ·  Generalized Linear Models (MA 575), Data Science in R (MA 415), Applied Statistics (MA 214)

✨ My Apps

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

🌲 Wilderness	✨ MBTI Vibe	🎬 What If Cinema	✉️ Letters from Screen	✈️ If You Disappeared	🎟️ Souvenirs
🗺️ Map of Me	🦋 A Room in Macondo	✒️ Say It Like a Classic	🏜️ Sanmao	🎨 Monet	🌻 Vincent

🎨 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