RFKG-CoT: Relation-Driven Adaptive Hop-count Selection and Few-Shot Path Guidance for Knowledge-Aware QA
This addresses reliability issues in QA for users of LLMs, but it is incremental as it builds on existing KG-CoT methods.
The paper tackles the problem of hallucinations in knowledge-intensive QA by large language models, proposing RFKG-CoT, which improves accuracy by up to 14.7 percentage points over a baseline on benchmarks like WebQSP.
Large language models (LLMs) often generate hallucinations in knowledge-intensive QA due to parametric knowledge limitations. While existing methods like KG-CoT improve reliability by integrating knowledge graph (KG) paths, they suffer from rigid hop-count selection (solely question-driven) and underutilization of reasoning paths (lack of guidance). To address this, we propose RFKG-CoT: First, it replaces the rigid hop-count selector with a relation-driven adaptive hop-count selector that dynamically adjusts reasoning steps by activating KG relations (e.g., 1-hop for direct "brother" relations, 2-hop for indirect "father-son" chains), formalized via a relation mask. Second, it introduces a few-shot in-context learning path guidance mechanism with CoT (think) that constructs examples in a "question-paths-answer" format to enhance LLMs' ability to understand reasoning paths. Experiments on four KGQA benchmarks show RFKG-CoT improves accuracy by up to 14.7 pp (Llama2-7B on WebQSP) over KG-CoT. Ablations confirm the hop-count selector and the path prompt are complementary, jointly transforming KG evidence into more faithful answers.