Logical reasoning with large language models (LLMs) has made significant progress in recent years. However, existing methods still suffer from insufficient rule semantic grounding and weak rule application mechanisms, making it difficult to achieve precise understanding and effective utilization of rules in complex multi-step reasoning. To address this, we propose Leibniz, a theory-of-mind driven neuro-symbolic reasoning framework. Specifically, we construct a bidirectional reasoning model based on multi-agent collaboration, which characterizes the reasoning process from two complementary perspectives, namely the Evolution Agent and the Reduction Agent. The former transforms belief-unstable propositions into stable ones that are beneficial for proving the target conclusion. The latter performs reverse reduction from the target to resolve belief conflicts and distill new inferential insights. Both share a belief state space and achieve efficient collaborative reasoning through continual belief updating. We integrate natural language and symbolic representations throughout the reasoning process. The experimental results demonstrate that Leibniz surpasses existing state-of-the-art models in reasoning accuracy, and further analyses reveal its substantial advantages in reliability and flexibility.

Driven by vast and diverse textual data, large language models (LLMs) have demonstrated impressive performance across numerous natural language processing (NLP) tasks. Yet, a critical question persists: does their generalization arise from mere memorization of training data or from deep semantic understanding? To investigate this, we propose a bi-perspective evaluation framework to assess LLMs’ scenario cognition—the ability to link semantic scenario elements with their arguments in context. Specifically, we introduce a novel scenario-based dataset comprising diverse textual descriptions of fictional facts, annotated with scenario elements. LLMs are evaluated through their capacity to answer scenario-related questions (model output perspective) and via probing their internal representations for encoded scenario elements-argument associations (internal representation perspective). Our experiments reveal that current LLMs predominantly rely on superficial memorization, failing to achieve robust semantic scenario cognition, even in simple cases. These findings expose critical limitations in LLMs’ semantic understanding and offer cognitive insights for advancing their capabilities.

Event Causality Identification (ECI) aims to identify fine-grained causal relationships between events in an unstructured text. Existing ECI methods primarily rely on knowledge enhanced and graph-based reasoning approaches, but they often overlook the dependencies between similar events. Additionally, the connection between unstructured text and structured knowledge is relatively weak. Therefore, this paper proposes an ECI method enhanced by LLM Knowledge and Concept-Level Event Relations (LKCER). Specifically, LKCER constructs a conceptual-level heterogeneous event graph by leveraging the local contextual information of related event mentions, generating a more comprehensive global semantic representation of event concepts. At the same time, the knowledge generated by COMET is filtered and enriched using LLM, strengthening the associations between event pairs and knowledge. Finally, the joint event conceptual representation and knowledge-enhanced event representation are used to uncover potential causal relationships between events. The experimental results show that our method outperforms previous state-of-the-art methods on both benchmarks, EventStoryLine and Causal-TimeBank.

The task of sequential model editing is to fix erroneous knowledge in Pre-trained Language Models (PLMs) efficiently, precisely and continuously. Although existing methods can deal with a small number of modifications, these methods experience a performance decline or require additional annotated data, when the number of edits increases. In this paper, we propose a Retrieval Augmented Sequential Model Editing framework (RASE) that leverages factual information to enhance editing generalization and to guide the identification of edits by retrieving related facts from the fact-patch memory we constructed. Our main findings are: (i) State-of-the-art models can hardly correct massive mistakes stably and efficiently; (ii) Even if we scale up to thousands of edits, RASE can significantly enhance editing generalization and maintain consistent performance and efficiency; (iii) RASE can edit large-scale PLMs and increase the performance of different editors. Moreover, it can integrate with ChatGPT and further improve performance. Our code and data are available at: https://github.com/sev777/RASE.