GraphRAG reasoning improved via reinforcement learning training

Original: GraphRAG still has a reasoning gap. This repo shows one way to train through it.

Source: x.com ↗