Embodied RAG++: Retrieval-Augmented Generation From What Exists to What Works in Embodied Scenes
Embodied RAG++: Retrieval-Augmented Generation From What Exists to What Works in Embodied Scenes
This paper introduces E2W-Bench, a benchmark linking dynamic hybrid scene graphs with real product manuals to evaluate capability-grounded robot-object assignment under constraints like availability and vertical reach. The authors also present Embodied RAG++, a spatially grounded retrieval framework that scopes retrieval to candidate object manuals and resolves feasibility through deterministic spatial reasoning. On E2W-Bench, Embodied RAG++ achieves a pair accuracy of 0.90 with Qwen2.5-32B and 1.00 with Llama3.3-70B, outperforming baselines which reached at most 0.30 and 0.55 respectively, while requiring the shortest preprocessing time among retrieval-based methods.
The authors introduce E2W-Bench and Embodied RAG++ to address capability-grounded robot-object assignment by integrating dynamic hybrid scene graphs with product manuals, reporting pair accuracy metrics on the benchmark.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
retrieval to the manuals linked to candidate objects, caches extracted capabilities per product model, and resolves robot--object feasibility through deterministic spatial reasoning. On E2W-Bench, Embodied RAG++ reaches a pair accuracy of 0.90 with Qwen2.5-32B and 1.00 with Llama3.3-70B, compared with at most 0.30 and 0.55 for the baselines, while requiring the shortest preprocessing time among the retrieval-based m
Open source S4
ethods and tying for the lowest total token count.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org