Mahdi Erfanian
Ph.D. Candidate at UIC | Research Intern at Microsoft
5453, Computer Design Research and Learning Center (CDRLC)
850 W Taylor street,
Chicago, IL 60607
| I’m Mahdi Erfanian, a Ph.D. candidate in Computer Science at the University of Illinois Chicago, where I am a member of the InDeXLab under the supervision of Dr. Abolfazl Asudeh. I received my M.S. in Computer Science from UIC (awarded en route to my Ph.D.) and my B.Sc. in Computer Engineering from Sharif University of Technology. I am currently a Ph.D. Research Intern with Microsoft’s **Code | AI** team, working on hallucination mitigation in large language models and generative AI coding agents. My research spans large language models, multimodal data management, generative AI, vector databases, information retrieval, and efficient AI systems. |
I build practical systems that use foundation models to address challenges in responsible data management and complex multimodal retrieval. My work has appeared in top-tier venues including VLDB, ICML, and the IEEE Data Engineering Bulletin. These systems include Needle, which achieves up to a 3x improvement in mean average precision over OpenAI’s CLIP on complex natural-language queries, and RSR, which accelerates binary and ternary matrix multiplication by up to 24x over NumPy and 2.5x on quantized LLMs.
news
| Apr 27, 2026 | Released BibTeX Verifier, an open-source, in-browser tool that checks .bib entries against CrossRef and Semantic Scholar—useful for catching metadata mistakes and AI-hallucinated citations. Everything runs locally; only titles are sent to public APIs. Live app · GitHub |
|---|---|
| Mar 20, 2026 | Serving as PC member and reviewer for top-tier venues!  PC Member for WWW 2026, CIKM 2026/2025/2024, KDD 2026, NeurIPS 2026. Reviewer for ICLR 2026, NeurIPS 2025 (DynaFront), TKDE 2025/2024, PETRA 2024 (ETHER-AI). |
| Mar 15, 2026 | Delivered guest lectures at UIC: CS516 (Responsible Data Science) on Generative AI and Fairness, and CS418 (Intro to Data Science) on Generative AI and Multimodal Data Management.  |
| Mar 01, 2026 | “NeedleDB: A Generative-AI Based System for Accurate and Efficient Image Retrieval using Complex NL Queries” has been submitted to VLDB 2026!  |
| Feb 01, 2026 | “Needle: A Generative AI-Powered Multi-modal Database for Answering Complex Natural Language Queries” has been submitted to KDD 2026!  |
| Jan 15, 2026 | Started a Ph.D. Research Internship at Microsoft (CodeAI team)!  Working on mitigating hallucination in LLMs and GenAI code agents including Copilot and Codex. |
| Dec 15, 2025 |  Received my M.S. in Computer Science from the University of Illinois Chicago, awarded en route to my Ph.D.! |
selected publications
- Delulu-FIM: A Verified Multilingual Benchmark for Code Hallucination Detection in Fill-in-the-Middle TasksarXiv preprint arXiv:2605.07024, 2026Submitted to EMNLP 2026
- Synthetic Hallucinations, Real Gains: Hard Negatives from Frontier Models for FIM Hallucination Mitigation2026Submitted to TMLR
- Generative AI for Multi-Modal Data ManagementIn VLDB 2026 Ph.D. Workshop, 2026Accepted for oral presentation
- Task-Aware Data Augmentation using Generative AI for Group-Distributional Robustness2025Submitted to The VLDB Journal
- NeedleDB: A Generative-AI-Based System for Accurate and Efficient Image Retrieval using Complex Natural Language Queries2026Preprint
- Needle: A Generative-AI-Powered Multimodal Database for Answering Complex Natural Language QueriesarXiv preprint arXiv:2412.00639, 2025Submitted to TMLR
- An Efficient Matrix Multiplication Algorithm for Accelerating Inference in Binary and Ternary Neural NetworksIn The 2025 International Conference on Machine Learning, 2025arXiv preprint arXiv:2411.06360
- Chameleon: Foundation Models for Fairness-Aware Multi-Modal Data Augmentation to Enhance Coverage of MinoritiesProceedings of the VLDB Endowment, 2024
-
- Coverage-based Data-centric Approaches for Responsible and Trustworthy AIBulletin of the IEEE Computer Society Technical Committee on Data Engineering, 2024