用合成数据蒸馏小模型，让机器人本地运行大模型能力
Distilling On-device Language Models for Robot Planning with Minimal Human Intervention
- 用自动生成任务和环境，从大模型获取计划并蒸馏成小模型
- 将Llama-3.2-3B性能提升至GPT-4o的93%以上，仅靠合成数据
- 支持多种机器人平台与室内外环境，几乎无需人工干预
大型语言模型（LLMs）为机器人提供了强大的上下文推理能力和自然人机接口。然而，当前基于LLM的机器人通常依赖云端模型，在通信不稳定的户外或工业环境中使用受限。我们提出PRISM框架，用于在设备端高效运行的小语言模型（SLM）机器人规划器的蒸馏，实现最小化人工监督。从现有LLM驱动的规划器出发，PRISM自动合成多样化任务与环境，从LLM获取规划方案，并利用该合成数据集蒸馏出紧凑的SLM，作为原模型的即插即用替代品。我们将PRISM应用于三个基于LLM的规划器：地图构建与探索、操作抓取及家庭协助任务。结果表明，仅使用合成数据，便使Llama-3.2-3B的性能从原先的10%-20%提升至超过93%的GPT-4o水平。此外，蒸馏后的规划器在异构机器人平台（地面与空中）及多样环境（室内与室外）间具有良好泛化能力。相关软件、训练模型与数据集已开源：https://zacravichandran.github.io/PRISM。
原文摘要 · Abstract (English)
Large language models (LLMs) provide robots with powerful contextual reasoning abilities and a natural human interface. Yet, current LLM-enabled robots typically depend on cloud-hosted models, limiting their usability in environments with unreliable communication infrastructure, such as outdoor or industrial settings. We present PRISM, a framework for distilling small language model (SLM)-enabled robot planners that run on-device with minimal human supervision. Starting from an existing LLM-enabled planner, PRISM automatically synthesizes diverse tasks and environments, elicits plans from the LLM, and uses this synthetic dataset to distill a compact SLM as a drop-in replacement of the source model. We apply PRISM to three LLM-enabled planners for mapping and exploration, manipulation, and household assistance, and we demonstrate that PRISM improves the performance of Llama-3.2-3B from 10-20% of GPT-4o's performance to over 93% - using only synthetic data. We further demonstrate that the distilled planners generalize across heterogeneous robotic platforms (ground and aerial) and diverse environments (indoor and outdoor). We release all software, trained models, and datasets at https://zacravichandran.github.io/PRISM.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。