0.76078 0.36863 0.98039S0.52941 0.53725 0.96471T0.29412 0.7098 0.94902a0.05882 0.87843 0.93333R \__color_backend_reset: \__color_backend_reset: \__color_backend_reset: \__color_backend_reset: : Scalable Task-Conditioned Retrieval for Long-Horizon Multimodal Robot Memory
Anonymous Authors, Mingfeng Yuan, Hao Zhang, Mahan Mohammadi, Runhao Li, Jinjun Shan, Steven L. Waslander · 2026
Read 0.76078 0.36863 0.98039S0.52941 0.53725 0.96471T0.29412 0.7098 0.94902a0.05882 0.87843 0.93333R \__color_backend_reset: \__color_backend_reset: \__color_backend_reset: \__color_backend_reset: : Scalable Task-Conditioned Retrieval for Long-Horizon Multimodal Robot Memory by Anonymous Authors, Mingfeng Yuan, Hao Zhang, Mahan Mohammadi, Runhao Li, Jinjun Shan, Steven L. Waslander (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-09255.txt.
Description
Mobile robots working long shifts need memory that backs planning, recall, and thought over open-ended orders at many levels of detail. Mingfeng Yuan and five co-authors present STaR, an agentic framework that builds a task-agnostic, multimodal long-term memory. It preserves fine object attributes, spatial relations, and dynamic events, then retrieves from it with a scalable task-conditioned algorithm based on the information bottleneck.
Tested on NaVQA campus scenes and WH-VQA, a warehouse benchmark with many look-alike objects built in Isaac Sim, STaR beats strong baselines with better win rates and far smaller position error. Deployment on a real Husky wheeled robot indoors and outdoors confirms robust long-horizon reasoning and practical utility for navigation.
Catalog record
| CREATOR | Anonymous Authors, Mingfeng Yuan, Hao Zhang, Mahan Mohammadi, Runhao Li, Jinjun Shan, Steven L. Waslander |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-09255