| title | EpiCon: Collective Agent Learning through Co-Evolving Multimodal Memory |
|---|---|
| authors | Ziyun Zeng, Hang Hua, Shaden Alshammari, Rogerio Feris, William T. Freeman, Jiebo Luo |
| date | 2026-09-29 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2609.37923 |
| upvotes | 9 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| primer | Memory for AI agents |
| abstract | Agents can learn from past executions, but enabling different agents to reuse and build on one another's experience remains challenging. We introduce EpiCon, a shared multimodal memory framework for agent collective learning without updating host model parameters. EpiCon links question-level memory evolution to a persistent experience bank through two independently trained 2B models: a memory controller and a tree self-organizer. The controller jointly refines textual guidance and visual evidence across attempts and selectively includes visual memory. The self-organizer consolidates lessons hierarchically and retrieves experience and rules for new problems. We evaluate EpiCon on eleven benchmarks spanning four multimodal task domains, using two harnesses and multiple backbones. A frozen bank improves other systems even with a single solving attempt. A second harness raises the original system's macro-average score by 2.6 points across eleven benchmarks. Across four host configurations, EpiCon improves macro-average scores by 1.7 to 4.9 points over No Memory and reduces memory-operation time by 67\% to 74\% relative to backbone-sized memory models. |
cite
@misc{zeng2026epicon,
title={EpiCon: Collective Agent Learning through Co-Evolving Multimodal Memory},
author={Ziyun Zeng and Hang Hua and Shaden Alshammari and Rogerio Feris and William T. Freeman and Jiebo Luo},
year={2026},
eprint={2609.37923},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2609.37923}
}