Abstract Large language models have achieved outstanding performance on various downstream tasks with their advanced understanding of natural language and zero-shot capability. However, they struggle with knowledge constraints, particularly in tasks requiring complex reasoning or extended logical sequences. These limitations can affect their performance in question answering by leading to inaccuracies and hallucinations. This paper proposes a novel framework called KnowledgeNavigator that leverages large language models on knowledge graphs to achieve accurate and interpretable multi-hop reasoning. Especially with an analysis-retrieval-reasoning process, KnowledgeNavigator searches the optimal path iteratively to retrieve external knowledge and guide the reasoning to reliable answers. KnowledgeNavigator treats knowledge graphs and large language models as flexible components that can be switched between different tasks without additional costs. Experiments on three benchmarks demonstrate that KnowledgeNavigator significantly improves the performance of large language models in question answering and outperforms all large language models-based baselines.
Large language model (LLM) has achieved outstanding performance on various downstream tasks with its powerful natural language understanding and zero-shot capability, but LLM still suffers from knowledge limitation. Especially in scenarios that require long logical chains or complex reasoning, the hallucination and knowledge limitation of LLM limit its performance in question answering (QA). In this paper, we propose a novel framework KnowledgeNavigator to address these challenges by efficiently and accurately retrieving external knowledge from knowledge graph and using it as a key factor to enhance LLM reasoning. Specifically, KnowledgeNavigator first mines and enhances the potential constraints of the given question to guide the reasoning. Then it retrieves and filters external knowledge that supports answering through iterative reasoning on knowledge graph with the guidance of LLM and the question. Finally, KnowledgeNavigator constructs the structured knowledge into effective prompts that are friendly to LLM to help its reasoning. We evaluate KnowledgeNavigator on multiple public KGQA benchmarks, the experiments show the framework has great effectiveness and generalization, outperforming previous knowledge graph enhanced LLM methods and is comparable to the fully supervised models.
Wheel-legged robots have fast and stable motion characteristics on flat roads, but there are the problems of poor balance ability and low movement level in special terrains such as rough roads. In this paper, a new type of wheel-legged robot with parallel four-bar mechanism is proposed, and the linear quadratic regulator (LQR) controller and fuzzy proportion differentiation (PD) jumping controller are designed and developed to achieve stable motion so that the robot has the ability to jump over obstacles and adapt to rough terrain. The amount of energy released by the parallel four-bar linkage mechanism changes with the change of the link angle, and the height of the jump trajectory changes accordingly, which improves the robot's ability to overcome obstacles facing vertical obstacles. Simulations and real scene tests are performed in different terrain environments to verify obstacle crossing capabilities. The simulation results show that, in the pothole terrain, the maximum height error of the two hip joint motors is 2 mm for the obstacle surmounting method of the adaptive retractable wheel-legs; in the process of single leg obstacle surmounting, the maximum height error of the hip joint motors is only 6.6 mm. The comparison of simulation data and real scene experimental results shows that the robot has better robustness in moving under complex terrains.
The feasibility of using a flow injection (FI) hydride generation technique in conjunction with atomic fluorescence spectrometry (AFS) was investigated. Parameters were established for the determination of Sb, As, Bi, Hg, Se and Te. Among the parameters that were found to have a more pronounced influence on performance were the concentration of NaBH4, the carrier gas flow, the observation height and the temperature of the atomizer cell. Compared with the manual sampling system (or batch system), the relative detection limits of the FI combination were better by factors of 2.5–10. By using FI, the sample volume was reduced to 500 mm3, hence, the absolute detection limits were even better with improvements of between 10- and 50-fold depending on the element. The absolute detection limit for Se using the FI technique was 0.035 ng, while with the batch system it was 0.8 ng. Similarly, Hg detection limits with the FI technique and the batch system were found to be 0.015 and 0.4 ng, respectively. The best improvement in the absolute detection limits was found for Te, which with the FI technique was 0.02 ng while with the batch system it was 1.0 ng. The linear ranges were typically 2–3 orders of magnitude of analyte concentrations, which is much wider than that of atomic absorption spectrometry. Sampling frequency was typically 120 injections per hour, and since a double-channel AFS instrument was used in this work, pairs of elements such as As, Sb and Bi, Hg were measured simultaneously, which equates to 240 measurements per hour. The technique was applied to the determination of hydride forming elements in geological reference materials.
Retrieving external knowledge and prompting large language models with relevant information is an effective paradigm to enhance the performance of question-answering tasks. Previous research typically handles paragraphs from external documents in isolation, resulting in a lack of context and ambiguous references, particularly in multi-document and complex tasks. To overcome these challenges, we propose a new retrieval framework IIER, that leverages Inter-chunk Interactions to Enhance Retrieval. This framework captures the internal connections between document chunks by considering three types of interactions: structural, keyword, and semantic. We then construct a unified Chunk-Interaction Graph to represent all external documents comprehensively. Additionally, we design a graph-based evidence chain retriever that utilizes previous paths and chunk interactions to guide the retrieval process. It identifies multiple seed nodes based on the target question and iteratively searches for relevant chunks to gather supporting evidence. This retrieval process refines the context and reasoning chain, aiding the large language model in reasoning and answer generation. Extensive experiments demonstrate that IIER outperforms strong baselines across four datasets, highlighting its effectiveness in improving retrieval and reasoning capabilities.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
