RSF-GLLM: Bridging the Semantic Gap in Multi-Hop Knowledge Graph QA via Recurrent Soft-Flow and Decoupled LLM Generation

Multi-hop Question Answering over Knowledge Graphs faces a critical challenge: traditional retrieve-then-read pipelines break differentiability, preventing the retriever from learning to bridge the semantic gap where intermediate nodes lack lexical overlap with the query. To address this, we propose RSF-GLLM, a framework decoupling differentiable graph reasoning from answer generation. Our Recurrent Soft-Flow (RSF) module employs a GRU-guided query updater to propagate continuous relevance scores, utilizing a dynamic gating mechanism to traverse semantically dissimilar bridge nodes via structural cues. We introduce flow sparsity regularization to theoretically guarantee convergence from soft probabilities to discrete reasoning paths. These paths are extracted and textualized to fine-tune a Large Language Model (LLM), ensuring generation is grounded in factual topology. Experiments on WebQSP and CWQ demonstrate that RSF-GLLM achieves competitive performance with superior inference efficiency compared to LLM based computationally expensive approaches.

Imagine asking an AI: "What awards did the director of Inception win?" To answer, the system must realize the director is Christopher Nolan, then look up his awards — a two-step journey through a database of facts. The catch: nothing in the question says "Christopher Nolan," so word-matching gets stuck at this hidden bridge, and the usual workaround — calling a language model many times per question — is slow and costly. We built RSF-GLLM, which separates reasoning from writing: a small, fast module walks the fact database by following relationships rather than words, while a focusing rule pushes it toward a single confident route, and a language model composes the answer from the extracted path. On standard benchmarks, RSF-GLLM is competitive with far more expensive systems while needing only one language-model call. Every answer carries the chain of facts behind it, letting users verify the reasoning instead of trusting a black box.