Large Language Model (LLM) agents have established a promising paradigm for Knowledge Graph Question Answering (KGQA). However, their practical performance is severely constrained by the pervasive incompleteness of real-world knowledge graphs (KGs). To address this, Incomplete KGQA (IKGQA) has emerged as a vital task, yet existing multi-agent frameworks often suffer from a logic-misalignment effect, where passive collaboration leads to retrieval failure and hallucination propagation when encountering structural gaps. In this paper, we propose IDLC, an Intent-Driven Logical Compensation framework designed for faithful IKGQA. Unlike traditional reactive debate, IDLC formalizes reasoning as a proactive deliberation task within a Dec-POMDP. It introduces an Active KG Explorer that generates explicit Gap Descriptors to translate structural gaps into actionable retrieval goals, alongside a Logical Consistency Verifier that performs bi-directional anchoring to intercept step-wise semantic drift. The entire multi-agent policy is jointly optimized via reinforcement learning with a specialized logic-fidelity reward, aligning agent behaviors toward faithful reasoning paths. Extensive experiments on BRINK and IKG-CWQ/WebQSP benchmarks demonstrate that IDLC outperforms state-of-the-art baselines in both reasoning performance and step-wise logical fidelity. Furthermore, IDLC exhibits robust zero-shot transferability across various incompleteness levels while maintaining computational efficiency, confirming the effectiveness of intent-driven coordination in repairing fractured reasoning chains.

Large language models (LLMs) excel in natural language tasks, with Chain-of-Thought (CoT) prompting enhancing reasoning through step-by-step decomposition. However, CoT struggles in knowledge-intensive tasks with multiple entities and implicit multi-hop relations, failing to connect entities systematically in zero-shot settings. Existing knowledge graph methods, limited by static structures, lack adaptability in complex scenarios. We propose DS-MHP, a zero-shot framework to enhance LLM reasoning in multi-entity relation tasks. DS-MHP operates in three stages: 1) constructing query-specific subgraphs by extracting entities and relations; 2) generating and refining multi-hop paths using a hybrid strategy of Breadth-First Search, greedy expansion, and LLM supplementation; and 3) guiding LLMs with subgraphs and paths, aggregating answers via majority voting. Evaluated on 12 datasets spanning commonsense, logical, symbolic, and arithmetic reasoning, DS-MHP outperforms baselines and state-of-the-art methods in nearly all benchmarks. It achieves overall average accuracy increases of 3.9% on Mistral-7B and 3.6% on GPT-3.5 Turbo compared to SOTA, with significant gains in logical and symbolic reasoning. Additionally, DS-MHP reduces runtime and LLM calls compared to SOTA, enhancing computational efficiency. These improvements demonstrate DS-MHP’s superior reasoning accuracy, explainability, and efficiency in complex multi-entity tasks.