Reproduce “Multimodal Filesystem Memory For Long-Horizon Embodied Agents” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Chloe Su, Jiahan Li, Zhenting Qi, Chaoran Cheng, Haoyu Chen, Sejoon Jun, Weirui Ye, Yilun Du

MFM lets long-horizon agents retrieve past visual experience by keeping episodes and indexing them with searchable contextual cues. Agents use the index to find relevant records, inspect frames, and revise notes or procedures without changing evidence. In tested settings, GPT-6-Astra averaged 16.0 of 24 Ender Dragon milestones and 0.440 mean reward on KiCad, but did not complete the Dragon task.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.