ManimAgent: Self-Evolving Multimodal Agents for Visual Education
EMNLP 2026 · Poster
A multimodal agent that generates Manim animations from scientific content and learns across tasks through a dual-channel memory bank of successful examples and validated failure patterns, without updating model weights.
BibTeX
@misc{jiang2026manimagentselfevolvingmultimodalagents,
title = {ManimAgent: Self-Evolving Multimodal Agents for Visual Education},
author = {Wenjia Jiang and Zongyuan Cai and Yuanhang Shao and Chenru Wang and Boyan Han and Zhixue Song and Keyu Chen and Shengwei An and Xu Yang and Zhou Yang},
year = {2026},
eprint = {2606.30296},
archivePrefix = {arXiv},
primaryClass = {cs.AI},
url = {https://arxiv.org/abs/2606.30296}
}