arXiv Robotics — research abstracts· Maria Attarian, Yifan Ruan, Claas Voelcker, Chenyang Cao, Amy Zhang, Peter Stone, Nicholas Rhinehart, Igor Gilitschenski·· 8 hours agoEditorial score28
A Guide to Training Data Adaptation Methods
Post-Training, Finetuning, and Steering:A Field Guide for Curating Training Data
Summary
Study compares adaptation techniques for robotics tasks. Critic-based steering outperforms supervised finetuning with failure data but lags behind full finetuning in performance.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source:arXiv Robotics — research abstracts · arxiv.org