Metric-Gradient Projection for Stable Multi-Agent Policy Learning

In three linesHPML (Hodge-Projected Multi-agent Learning) stabilizes multi-agent learning by projecting the joint update field onto a metric-gradient component. The method uses Hodge-type projection in an L² space of vector fields, implemented via graph-based and amortized neural realizations. Results: improved stability and normalized returns on CTDE benchmarks.

Summary generated by Claude — human-verified