AdaPath: Query-Adaptive Path-Finding via Path-Bank for Multi-Hop Implicit Biomedical KGQA
This episode discusses the paper "AdaPath," a path-finding framework designed for complex, multi-hop implicit biomedical questions. The hosts explain how AdaPath overcomes challenges like elusive guidance cues and spurious paths by using a Path-Bank to manage knowledge. It prunes irrelevant data and consistently outperforms baselines, enabling AI to follow causal chains in biology.