Vision Language Models - Connect Session
Details
Announcing the first ResearchTrend.AI VLM Connect Session in 2026!
We are back and energized to dive into the latest in Vision-Language Models. To start the year right, we have two fascinating talks covering Temporal Medical AI and High-Efficiency Retrieval.
Agenda (UTC) - Monday, January 19th
07:00 - 07:30: Junyi Zhang
📄 TemMed-Bench: Evaluating Temporal Medical Image Reasoning 💡 The Insight: Real doctors track patient history, but most VLMs don't. Junyi introduces TemMed-Bench, the first benchmark designed to test if models can reason about changes in patient conditions over time, revealing why multi-modal retrieval is the key to future medical AI.
07:30 - 08:00: Paul Teiletche
📄 ModernVBERT: Towards Smaller Visual Document Retrievers 💡 The Insight: Bigger isn't always better. Paul presents ModernVBERT, a compact 250M-parameter encoder. By rethinking the training pipeline, it outperforms models 10x its size on document retrieval tasks, making RAG systems faster and more accessible on standard hardware.
🌟 Start your 2026 research journey with us!