Reproduce “Leave It Better Than Found: Value-Guided Cooperation in Sequential Tasks” 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
Mario Gomez Andreu, Michael Amir, Antonio Marino, Tianxu An, Marco Hutter, Amanda Prorok
PAVE lets a leader use a follower’s learned value function to choose a useful handoff state, without sharing the follower’s policy. It combines that preference with the leader’s objective and discounts estimates for states unlike the follower’s training data. In simulated Warehouse Pickup Blocked, follower success was 89.94 ± 10.02%, versus 49.78 ± 10.26% for Naive; experiments cover two ordered agents.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.