MarsTSC
Empowering VLMs for Few-Shot Multimodal Time Series Classification via Tailored Agentic Reasoning
MarsTSC equips vision-language models with a self-evolving knowledge bank for few-shot time-series classification. A Generator, Reflector, and Modifier work together to explain decisions, diagnose errors, and preserve verified insights across new samples.
Read paperFramework overview: multimodal feature discovery, reflective two-pass reasoning, and continuous knowledge-bank refinement.
Figure from the open paper · arXiv:2605.09395