LIBERO-PeRM: Benchmarking Personalized Robotic Manipulation
LIBERO-PeRM benchmarks whether robot policies satisfy user preferences separately from completing manipulation tasks.
Researchers introduce LIBERO-PeRM, a simulation benchmark that separates task completion from preference satisfaction for personalized robotic manipulation. It covers language-expressible motion, skill, and user-profile preferences across five suites totaling 312 tasks with paired demonstrations. Experiments find current vision-language-action policies struggle at zero-shot personalization, while supervised fine-tuning makes preferences learnable. Preference generalization, decomposition, and conflict handling remain difficult.
- Benchmark separates task completion from preference satisfaction.
- Five suites contain 312 tasks with paired demonstrations.
- Preferences cover motion, skill, and user-profile choices.
- Current VLA policies struggle with zero-shot personalization.
- Generalization, decomposition, and conflict handling remain difficult.
From the original article
1 min read at arxiv.orgPersonalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulation has been extensively studied, personalized robotic manipulation (PeRM) lacks a controlled benchmark for separating task completion from preference satisfaction. We introduce LIBERO-PeRM, a benchmark for preference learning in embodied agents.
By Zhixu Li · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.