TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning
The paper introduces TaskGround, a framework designed for structured executable task inference in household settings. It addresses the challenges of operating from complete household scenes and situated requests, which often contain irrelevant information. TaskGround improves task success rates significantly and enhances the effectiveness of compact local models for practical household deployment.
- ▪TaskGround is a training-free and model-agnostic framework for household reasoning.
- ▪It infers executable task structures from complete household scenes and situated requests.
- ▪The framework has been evaluated using FullHome, an evaluation suite of 400 household tasks.
- ▪TaskGround improves task success rates and reduces input-token costs significantly.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.18109 (cs) [Submitted on 18 May 2026] Title:TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning Authors:ZhiYuan Feng, Yu Deng, Ruichuan An, Zhenhua Liu, Qixiu Li, Keming Wu, Zhiying Du, Weijie Wang, Haoxiao Wang, Shuang Chen, Sicheng Xu, Yaobo Liang, Jiaolong Yang, Baining Guo View a PDF of the paper titled TaskGround: Structured Executable Task Inference for Full-Scene Household Reasoning, by ZhiYuan Feng and 13 other authors View PDF HTML (experimental) Abstract:In real home deployments, household agents must often operate from a complete household scene and a situated household request, rather than from a clean task specification.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
