Tianzhe Zhao

Fangzhi Xu, Qika Lin, Jiawei Han et al.

Logical reasoning consistently plays a fundamental and significant role in the domains of knowledge engineering and artificial intelligence. Recently, Large Language Models (LLMs) have emerged as a noteworthy innovation in natural language processing (NLP). However, the question of whether LLMs can effectively address the task of logical reasoning, which requires gradual cognitive inference similar to human intelligence, remains unanswered. To this end, we aim to bridge this gap and provide comprehensive evaluations in this paper. Firstly, to offer systematic evaluations, we select fifteen typical logical reasoning datasets and organize them into deductive, inductive, abductive and mixed-form reasoning settings. Considering the comprehensiveness of evaluations, we include 3 early-era representative LLMs and 4 trending LLMs. Secondly, different from previous evaluations relying only on simple metrics (e.g., \emph{accuracy}), we propose fine-level evaluations in objective and subjective manners, covering both answers and explanations, including \emph{answer correctness}, \emph{explain correctness}, \emph{explain completeness} and \emph{explain redundancy}. Additionally, to uncover the logical flaws of LLMs, problematic cases will be attributed to five error types from two dimensions, i.e., \emph{evidence selection process} and \emph{reasoning process}. Thirdly, to avoid the influences of knowledge bias and concentrate purely on benchmarking the logical reasoning capability of LLMs, we propose a new dataset with neutral content. Based on the in-depth evaluations, this paper finally forms a general evaluation scheme of logical reasoning capability from six dimensions (i.e., \emph{Correct}, \emph{Rigorous}, \emph{Self-aware}, \emph{Active}, \emph{Oriented} and \emph{No hallucination}). It reflects the pros and cons of LLMs and gives guiding directions for future works.

Fangzhi Xu, Jun Liu, Qika Lin et al.

Logical reasoning task involves diverse types of complex reasoning over text, based on the form of multiple-choice question answering. Given the context, question and a set of options as the input, previous methods achieve superior performances on the full-data setting. However, the current benchmark dataset has the ideal assumption that the reasoning type distribution on the train split is close to the test split, which is inconsistent with many real application scenarios. To address it, there remain two problems to be studied: (1) How is the zero-shot capability of the models (train on seen types and test on unseen types)? (2) How to enhance the perception of reasoning types for the models? For problem 1, we propose a new benchmark for generalized zero-shot logical reasoning, named ZsLR. It includes six splits based on the three type sampling strategies. For problem 2, a type-aware model TaCo is proposed. It utilizes both the heuristic input reconstruction and the contrastive learning to improve the type perception in the global representation. Extensive experiments on both the zero-shot and full-data settings prove the superiority of TaCo over the state-of-the-art methods. Also, we experiment and verify the generalization capability of TaCo on other logical reasoning dataset.

Tianzhe Zhao, Jiaoyan Chen, Shuxiu Zhang et al.

Large language models (LLMs) have achieved remarkable success across a wide range of applications especially when augmented by external knowledge through retrieval-augmented generation (RAG). Despite their widespread adoption, recent studies have shown that LLMs often struggle to perform faithful reasoning when conflicting knowledge is retrieved. However, existing work primarily focuses on conflicts between external knowledge and the parametric knowledge of LLMs, leaving conflicts across external knowledge largely unexplored. Meanwhile, modern RAG systems increasingly emphasize the integration of unstructured text and (semi-)structured data like knowledge graphs (KGs) to improve knowledge completeness and reasoning faithfulness. To address this gap, we introduce ConflictQA, a novel benchmark that systematically instantiates conflicts between textual evidence and KG evidence. Extensive evaluations across representative LLMs reveal that, facing such cross-source conflicts, LLMs often fail to identify reliable evidence for correct reasoning. Instead, LLMs become more sensitive to prompting choices and tend to rely exclusively on either KG or textual evidence, resulting in incorrect responses. Based on these findings, we further propose XoT, a two-stage explanation-based thinking framework tailored for reasoning over heterogeneous conflicting evidence, and verify its effectiveness with extensive experiments.

Rongman Xu, Yifei Li, Tianzhe Zhao et al.

Large Language Models (LLMs) have demonstrated remarkable abilities in reasoning. However, maximizing their potential through inference-time scaling faces challenges in trade-off between sampling budget and reasoning quality. Current strategies remain inefficient as they typically treat sampling width and depth as orthogonal objectives, where width consensus methods risk reinforcing hallucinations, while depth pruning mechanisms prematurely truncate complex yet valid reasoning chains. Therefore, we propose Dual-Dimensional Consistency (DDC), a unified framework that bridges path quality with adaptive termination. By coupling Confidence-Weighted Bayesian protocol with a Trend-Aware Stratified Pruning, our method ensures that computational resources are concentrated on high quality reasoning paths, filtering hallucinations while accelerating consensus. Evaluations across five benchmarks demonstrate that this approach reduces token consumption by over 10 times while maintaining or exceeding the accuracy of strong baselines across various LLMs.

Yifei Li, Lingling Zhang, Hang Yan et al.

Traditional knowledge graph (KG) embedding methods aim to represent entities and relations in a low-dimensional space, primarily focusing on static graphs. However, real-world KGs are dynamically evolving with the constant addition of entities, relations and facts. To address such dynamic nature of KGs, several continual knowledge graph embedding (CKGE) methods have been developed to efficiently update KG embeddings to accommodate new facts while maintaining learned knowledge. As KGs grow at different rates and scales in real-world scenarios, existing CKGE methods often fail to consider the varying scales of updates and lack systematic evaluation throughout the entire update process. In this paper, we propose SAGE, a scale-aware gradual evolution framework for CKGE. Specifically, SAGE firstly determine the embedding dimensions based on the update scales and expand the embedding space accordingly. The Dynamic Distillation mechanism is further employed to balance the preservation of learned knowledge and the incorporation of new facts. We conduct extensive experiments on seven benchmarks, and the results show that SAGE consistently outperforms existing baselines, with a notable improvement of 1.38% in MRR, 1.25% in H@1 and 1.6% in H@10. Furthermore, experiments comparing SAGE with methods using fixed embedding dimensions show that SAGE achieves optimal performance on every snapshot, demonstrating the importance of adaptive embedding dimensions in CKGE. The codes of SAGE are publicly available at: https://github.com/lyfxjtu/Dynamic-Embedding.

Qika Lin, Tianzhe Zhao, Kai He et al.

Due to the presence of the natural gap between Knowledge Graph (KG) structures and the natural language, the effective integration of holistic structural information of KGs with Large Language Models (LLMs) has emerged as a significant question. To this end, we propose a two-stage framework to learn and apply quantized codes for each entity, aiming for the seamless integration of KGs with LLMs. Firstly, a self-supervised quantized representation (SSQR) method is proposed to compress both KG structural and semantic knowledge into discrete codes (\ie, tokens) that align the format of language sentences. We further design KG instruction-following data by viewing these learned codes as features to directly input to LLMs, thereby achieving seamless integration. The experiment results demonstrate that SSQR outperforms existing unsupervised quantized methods, producing more distinguishable codes. Further, the fine-tuned LLaMA2 and LLaMA3.1 also have superior performance on KG link prediction and triple classification tasks, utilizing only 16 tokens per entity instead of thousands in conventional prompting methods.

Tianzhe Zhao, Jiaoyan Chen, Yanchi Ru et al.

Knowledge graph embedding (KGE) methods have achieved great success in handling various knowledge graph (KG) downstream tasks. However, KGE methods may learn biased representations on low-quality KGs that are prevalent in the real world. Some recent studies propose adversarial attacks to investigate the vulnerabilities of KGE methods, but their attackers are target-oriented with the KGE method and the target triples to predict are given in advance, which lacks practicability. In this work, we explore untargeted attacks with the aim of reducing the global performances of KGE methods over a set of unknown test triples and conducting systematic analyses on KGE robustness. Considering logic rules can effectively summarize the global structure of a KG, we develop rule-based attack strategies to enhance the attack efficiency. In particular,we consider adversarial deletion which learns rules, applying the rules to score triple importance and delete important triples, and adversarial addition which corrupts the learned rules and applies them for negative triples as perturbations. Extensive experiments on two datasets over three representative classes of KGE methods demonstrate the effectiveness of our proposed untargeted attacks in diminishing the link prediction results. And we also find that different KGE methods exhibit different robustness to untargeted attacks. For example, the robustness of methods engaged with graph neural networks and logic rules depends on the density of the graph. But rule-based methods like NCRL are easily affected by adversarial addition attacks to capture negative rules

Qika Lin, Yifan Zhu, Bin Pu et al.

Medical foundation models (FMs) have shown tremendous promise amid the rapid advancements in artificial intelligence (AI) technologies. However, current medical FMs typically generate answers in a black-box manner, lacking transparent reasoning processes and locally grounded interpretability, which hinders their practical clinical deployments. To this end, we introduce DeepMedix-R1, a holistic medical FM for chest X-ray (CXR) interpretation. It leverages a sequential training pipeline: initially fine-tuned on curated CXR instruction data to equip with fundamental CXR interpretation capabilities, then exposed to high-quality synthetic reasoning samples to enable cold-start reasoning, and finally refined via online reinforcement learning to enhance both grounded reasoning quality and generation performance. Thus, the model produces both an answer and reasoning steps tied to the image's local regions for each query. Quantitative evaluation demonstrates substantial improvements in report generation (e.g., 14.54% and 31.32% over LLaVA-Rad and MedGemma) and visual question answering (e.g., 57.75% and 23.06% over MedGemma and CheXagent) tasks. To facilitate robust assessment, we propose Report Arena, a benchmarking framework using advanced language models to evaluate answer quality, further highlighting the superiority of DeepMedix-R1. Expert review of generated reasoning steps reveals greater interpretability and clinical plausibility compared to the established Qwen2.5-VL-7B model (0.7416 vs. 0.2584 overall preference). Collectively, our work advances medical FM development toward holistic, transparent, and clinically actionable modeling for CXR interpretation.

Tianzhe Zhao, Jiaoyan Chen, Yanchi Ru et al.

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by retrieving external data to mitigate hallucinations and outdated knowledge issues. Benefiting from the strong ability in facilitating diverse data sources and supporting faithful reasoning, knowledge graphs (KGs) have been increasingly adopted in RAG systems, giving rise to KG-based RAG (KG-RAG) methods. Though RAG systems are widely applied in various applications, recent studies have also revealed its vulnerabilities to data poisoning attacks, where malicious information injected into external knowledge sources can mislead the system into producing incorrect or harmful responses. However, these studies focus exclusively on RAG systems using unstructured textual data sources, leaving the security risks of KG-RAG largely unexplored, despite the fact that KGs present unique vulnerabilities due to their structured and editable nature. In this work, we conduct the first systematic investigation of the security issue of KG-RAG methods through data poisoning attacks. To this end, we introduce a practical, stealthy attack setting that aligns with real-world implementation. We propose an attack strategy that first identifies adversarial target answers and then inserts perturbation triples to complete misleading inference chains in the KG, increasing the likelihood that KG-RAG methods retrieve and rely on these perturbations during generation. Through extensive experiments on two benchmarks and four recent KG-RAG methods, our attack strategy demonstrates strong effectiveness in degrading KG-RAG performance, even with minimal KG perturbations. In-depth analyses are also conducted to understand the safety threats within the internal stages of KG-RAG systems and to explore the robustness of LLMs against adversarial knowledge.

Fangzhi Xu, Qika Lin, Jun Liu et al.

Textbook Question Answering (TQA) is a complex multimodal task to infer answers given large context descriptions and abundant diagrams. Compared with Visual Question Answering (VQA), TQA contains a large number of uncommon terminologies and various diagram inputs. It brings new challenges to the representation capability of language model for domain-specific spans. And it also pushes the multimodal fusion to a more complex level. To tackle the above issues, we propose a novel model named MoCA, which incorporates multi-stage domain pretraining and multimodal cross attention for the TQA task. Firstly, we introduce a multi-stage domain pretraining module to conduct unsupervised post-pretraining with the span mask strategy and supervised pre-finetune. Especially for domain post-pretraining, we propose a heuristic generation algorithm to employ the terminology corpus. Secondly, to fully consider the rich inputs of context and diagrams, we propose cross-guided multimodal attention to update the features of text, question diagram and instructional diagram based on a progressive strategy. Further, a dual gating mechanism is adopted to improve the model ensemble. The experimental results show the superiority of our model, which outperforms the state-of-the-art methods by 2.21% and 2.43% for validation and test split respectively.

Yudai Pan, Jun Liu, Lingling Zhang et al.

Relation reasoning in knowledge graphs (KGs) aims at predicting missing relations in incomplete triples, whereas the dominant paradigm is learning the embeddings of relations and entities, which is limited to a transductive setting and has restriction on processing unseen entities in an inductive situation. Previous inductive methods are scalable and consume less resource. They utilize the structure of entities and triples in subgraphs to own inductive ability. However, in order to obtain better reasoning results, the model should acquire entity-independent relational semantics in latent rules and solve the deficient supervision caused by scarcity of rules in subgraphs. To address these issues, we propose a novel graph convolutional network (GCN)-based approach for interpretable inductive reasoning with relational path contrast, named RPC-IR. RPC-IR firstly extracts relational paths between two entities and learns representations of them, and then innovatively introduces a contrastive strategy by constructing positive and negative relational paths. A joint training strategy considering both supervised and contrastive information is also proposed. Comprehensive experiments on three inductive datasets show that RPC-IR achieves outstanding performance comparing with the latest inductive reasoning methods and could explicitly represent logical rules for interpretability.