HuggingFace Daily Papers（社区热门论文）· HuggingFace Daily Papers（社区热门论文）·· 7 天前AI 评分41
EpiCon：通过共同演化多模态记忆实现智能体集体学习
AI 导读
EpiCon 是一个无需更新宿主模型参数的多模态共享记忆框架，通过两个独立训练的 2B 模型——记忆控制器与树状自组织器——连接问题级记忆演化与持久经验库，实现不同智能体之间的经验复用。在覆盖四个多模态任务域的 11 个基准上，第二个 harness 将原系统宏平均分提升 2.6 分；四种宿主配置下较无记忆基线提升 1.7 至 4.9 分，记忆操作时间减少 67% 至 74%。
正文
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.
| Comments: | Preprint |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2609.37923 [cs.CV] |
| (or arXiv:2609.37923v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2609.37923 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ziyun Zeng [view email]
[v1]
Tue, 29 Sep 2026 16:16:24 UTC (3,296 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org