#001 · Research note · Jul 7, 2026
JEPA: From Language Models to World Models
Why JEPA’s latent-prediction objective may shift AI systems from token prediction toward predictive world models for VLMs, VLAs, and embodied agents.
Research at the edge of video, vision-language models, and agents.
Research projects, papers, preprints, talks, and references from the VideoDB research archive.
All /
Research
Research
#002 · Research note · May 15, 2026
How to Evaluate Multimodal VLMs for Your Video Use Case
A practical workflow for evaluating video VLM setups with VideoDB and Langfuse, from task definition and dataset design to tracing, scoring, and deployment decisions.
#003 · arXiv preprint arXiv:2604.11177 · 2026/4/13
Do Thought Streams Matter? Evaluating Reasoning in Gemini Vision-Language Models for Video Scene Understanding
Benchmarks how internal reasoning traces affect video scene understanding in Gemini models, including where quality gains plateau and how tight budgets increase compression-step hallucination.
#004 · arXiv preprint arXiv:2502.06445 · 2025/2/10
Benchmarking Vision-Language Models on Optical Character Recognition in Dynamic Video Environments
Introduces an open-source benchmark for evaluating VLMs on OCR tasks in dynamic video environments across 1,477 manually annotated frames.
#005 · Systems 9(3), 53 · 2021/7/13
Modeling the Spread and Control of COVID-19
Presents an agent-based extension of epidemic modeling for confined spaces, comparing black-box and glass-box views while simulating lockdowns, social distancing, hygiene, quarantine, and hospitalization policies.
#006 · IEEE Transactions on Computational Social Systems 5(1), 277–288 · 2018/1/17
Agent-Based Modeling of Emergency Evacuations Considering Human Panic Behavior
Models emergency evacuations with agents whose panic behavior is shaped by psychological and physical factors, then uses simulations to identify bottlenecks and compare evacuation strategies.
#007 · PyCon India talk · 2015
Solving Logical Puzzles with Natural Language Processing
A practical talk that starts from rule-based NLP features and WordNet, then motivates distributed word representations and deep learning for solving logical puzzles from natural language.
#008 · Research project · Ongoing
Chitrakāvya
Sanskrit picture-poems rebuilt as computational artifacts, connecting sloka, geometry, knight's tours, palindromes, and early algorithmic thinking.
#009 · Research project · Ongoing
Emergence
Notes on multi-agent systems, collective intelligence, alignment, swarm behavior, and computational models of consciousness.