Ehsan Latif
AI Research Scientist
I am an AI Research Scientist focusing on LLMs, multi‑agent systems, and trustworthy AI. My work has been recognized with a Best Paper Nomination at AIED 2025, and published at top AI conferences such as NeurIPS, AAAI, AIED, IROS, and ICRA.
I co‑organized and chaired workshops/sessions at AIED 2024 and AIED 2025, and lead research across NSF/IES‑funded projects on automated scoring and AI‑augmented education. I train and deploy transformer models (BERT, GPT, LLaMA) and design multi‑robot/agent algorithms for coordination and localization—bridging fundamental research with real‑world impact.
Selected Publications
SketchMind: A Multi-Agent Cognitive Framework for Assessing Student-Drawn Scientific Sketches New
Accepted to NeurIPS 2025; introduces a multi-agent cognitive framework for robust assessment of student-drawn scientific sketches.
Efficient multi-task inference: Model merging with Gromov-Wasserstein feature alignment New
Presents an efficient model-merging approach for multi-task inference via Gromov–Wasserstein feature alignment.
Latest News
Paper accepted at NeurIPS 2025: SketchMind
“SketchMind: A Multi-Agent Cognitive Framework for Assessing Student-Drawn Scientific Sketches” authored by Ehsan Latif, Z. Khan, & X. Zhai — accepted to NeurIPS 2025.
Read moreAIED 2025 Best Paper Nomination
“Efficient multi-task inference: Model merging with Gromov-Wasserstein feature alignment” authored by Ehsan Latif, L. Fang, H. Lu, Y. Zhou, P. Ma, & X. Zhai — nominated for Best Paper at AIED 2025.
Read moreTwo papers accepted at AIED 2025
Two publications at AIED 2025: (1) “Privacy-Preserved Automated Scoring using Federated Learning for Educational Research” authored by Ehsan Latif, X. Zhai; (2) “Efficient multi-task inference: Model merging with Gromov-Wasserstein feature alignment” authored by Ehsan Latif, L. Fang, H. Lu, Y. Zhou, P. Ma, & X. Zhai.
Read more