Tung Sum Thomas Kwok
I am a second-year PhD student at University of California Los Angeles and a member of the Trustworthy AI Lab advised by Professor Guang Cheng. My research interests lie on the large language models (LLMs), world models and tabular data modeling. Happy to connect and discuss further!
📢 News
- Sep 2026: 2 papers accepted to NeurIPS 2026.
- June 2026: Received the Most Promising Statistician Award from the UCLA Department of Statistics and Data Science.
- Feb 2026: Paper accepted to the Pacific-Asia Conference on Knowledge Discovery and Data Mining
- Jan 2026: Paper accepted to The ACM Web Conference 2026.
- Jan 2026: VeritasFi won Best Paper at AAAI 2026 Workshop on Agentic AI in Financial Services.
- Sep 2025: Began PhD studies at UCLA.
- July 2025: Paper accepted to the KDD’ 25 Workshop on Evaluation and Trustworthiness of Agentic and Generative AI Models.
- May 2025: Paper accepted to The Conference on Information and Knowledge Management 2025.
- Feb 2025: Paper accepted to ICDE’ 25 Data Engineering Meets Large Language Models: Challenges and Opportunities Workshop.
- Jan 2025: Graduated from Master’s of Mathematical Finance and Financial Technology program at Boston University.
đź“„ Selected Publications (full list)
PreprintLEDFlow: Introducing Entropy-Guided Generation Order into Uniform Discrete Flow
Training-free sampler that fixes low-entropy (reliable) token predictions first while letting uncertain positions keep evolving under uniform discrete flow
NeurIPS 2026From Table to Cell: Attention for Better Reasoning with TabAlign
Diffusion-denoised reasoning plans paired with cell-level attention grounding for stepwise table reasoning
NeurIPS 2026TabWorld: A World-Modeling Foundation Model for Tabular Generation
A generative tabular foundation model that learns a table-adaptive simulator, combining observational diffusion training with query-conditioned world-model supervision to predict responses to structural interventions
WWW 2026VeritasFi: Multi-tiered RAG for Multimodal Financial QA
Hybrid RAG framework with multi-modal preprocessing and retrieval components for financial documents
CIKM 2025FinSage: A Multi-aspect RAG System for Financial Filings Question Answering
A multi-path sparse-dense retrieval system augmented with query expansion, metadata-aware semantic search and a domain-specialized re-ranking module fine-tuned via Direct Preference Optimization
ICDE 2025 WorkshopGReaTER: Generate Realistic Tabular data after data Enhancement and Reduction
Tabular data synthesis with Semantic Enhancements
đź’Ľ Selected Experience
- Adaptive Investments — Quantitative Research Intern (2024): financial RL, option and bond pricings
- Hammer Capital — Full-time Investment Analyst (2021-2023): leveraged notes origination, debt restructuring advisory
✨ Honors & Awards
- Most Promising Statistician Award, UCLA Statistics and Data Science (2026)
- AAAI 2026 Workshop Best Paper Award (2026)
- OpenAI Researcher Access Program Award (2025)
- ICAIF 2024 Workshop Best Presentation Award (2024)
- Highest Honor Graduation, Boston University (2025)
- BU MSMFT Dean’s Achievement Scholarship (2023)
- CUHK Dean’s List Award (2020)
🌍 Academic Services
- Program committee/reviewer: ICML 2026, COLM 2026, NeurIPS 2026
- Conference student volunteer: ICLR 2026
- Panelist: ACM@UCLA 2025
- Teaching: Guest lecturer for STATS161/261 (Winter 2025, Fall 2025), STATS163/263 (Winter 2026); course assistant/grader for STATS10, STATS163/263
🧩 Miscellaneous
Former track athlete (10th in Hong Kong Junior Triple Jump Ranking 2016 - 2017; 6th in USFHK 2018 - 2019). Also into poker, Chinese chess and basketball.