Reproduce “Interpretability by Design for Efficient Multi-Objective Reinforcement Learning” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Qiyue Xia, Tianwei Wang, J. Michael Herrmann

LLE-MORL generates policies for multi-objective reinforcement learning by extending local parameter directions instead of training each candidate from scratch. It briefly retrains under nearby preferences, then evaluates and fine-tunes selected candidates. On MO-Hopper-2d, it reports Hypervolume of 3.02 ± 0.28 × 10⁵, versus 1.96 ± 0.19 × 10⁵ for MORL/D at equal interaction budget; this supports local gains, not recovery of every Pareto branch.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.