Large Language models (LLMs) are revolutionizing the conversational recommender systems (CRS) through their impressive capabilities in instruction comprehension, reasoning, and human interaction. A core factor underlying effective dialogue is the ability to infer and reason about others' mental states (such as desire, intention, and belief), a cognitive capacity commonly referred to as Theory of Mind (ToM). Despite growing interest in evaluating ToM in LLMs, current benchmarks predominantly rely on synthetic narratives inspired by Sally-Anne test, which emphasize physical perception and fail to capture the complexity of mental state inference in real-world conversational settings. Moreover,existing benchmarks often overlook a critical component of human ToM: behavioral prediction, the ability to use inferred mental states to guide strategic decision-making and select appropriate conversational actions for future interactions. To better align LLM-based ToM evaluation with human-like social reasoning, we propose RecToM, a novel benchmark for evaluating ToM abilities in recommendation dialogues. RecToM focuses on two complementary dimensions: Cognitive Inference and Behavioral Prediction. The former focus on understanding what has been communicated by inferring the underlying mental states. The latter emphasizes what should be done next, evaluating whether LLMs can leverage these inferred mental states to predict, select, and assess appropriate dialogue strategies. Together, these dimensions enable a comprehensive assessment of ToM reasoning in CRS. Extensive experiments on state-of-the-art LLMs demonstrate that RecToM poses a significant challenge. While the models exhibit partial competence in recognizing mental states, they struggle to maintain coherent, strategic ToM reasoning throughout dynamic recommendation dialogues, particularly in tracking evolving intentions and aligning conversational strategies with inferred mental states.

Theory of Mind (ToM), the ability to attribute mental states to others, is a hallmark of social intelligence.While large language models (LLMs) demonstrate promising performance on standard ToM benchmarks, we observe that they often fail to generalize to complex taskspecific scenarios, relying heavily on prompt scaffolding to mimic reasoning.The critical misalignment between the internal knowledge and external behavior raises a fundamental question: Do LLMs truly possess intrinsic cognition, and can they externalize this internal knowledge into stable, high-quality behaviors?To answer this, we introduce COSTOM 1 (Causal-oriented Steering for ToM alignment), a framework that transitions from mechanistic interpretation to active intervention.First, we employ causal tracing to map the internal distribution of ToM features, empirically uncovering the internal layers' characteristics in encoding fundamental ToM semantics.Building on this insight, we implement a lightweight alignment framework via targeted activation steering within these ToM-critical layers.Experiments demonstrate that COSTOM significantly enhances human-like social reasoning capabilities and downstream dialogue quality.

Abstract Knowledge Graphs (KGs) are pivotal for effectively organizing and managing structured information across various applications. Financial KGs have been successfully employed in advancing applications such as audit, anti-fraud, and anti-money laundering. Despite their success, the construction of Chinese financial KGs has seen limited research due to the complex semantics. A significant challenge is the overlap triples problem, where entities feature in multiple relations within a sentence, hampering extraction accuracy–more than 39% of the triples in Chinese datasets exhibit the overlap triples. To address this, we propose the Entity-type-Enriched Cascaded Neural Network (E 2 CNN), leveraging special tokens for entity boundaries and types. E 2 CNN ensures consistency in entity types and excludes specific relations, mitigating overlap triple problems and enhancing relation extraction. Besides, we introduce the available Chinese financial dataset F in C orpus .CN, annotated from annual reports of 2,000 companies, containing 48,389 entities and 23,368 triples. Experimental results on the DUIE dataset and F in C orpus .CN underscore E 2 CNN’s superiority over state-of-the-art models.

Knowledge graph (KG) representation learning aims to map entities and relations into a low-dimensional representation space, showing significant potential in many tasks. Existing approaches follow two categories: (1) Graph-based approaches encode KG elements into vectors using structural score functions. (2) Text-based approaches embed text descriptions of entities and relations via pre-trained language models (PLMs), further fine-tuned with triples. We argue that graph-based approaches struggle with sparse data, while text-based approaches face challenges with complex relations. To address these limitations, we propose a unified Text-Augmented Attention-based Recurrent Network, bridging the gap between graph and natural language. Specifically, we employ a graph attention network based on local influence weights to model local structural information and utilize a PLM based prompt learning to learn textual information, enhanced by a mask-reconstruction strategy based on global influence weights and textual contrastive learning for improved robustness and generalizability. Besides, to effectively model multi-hop relations, we propose a novel semantic-depth guided path extraction algorithm and integrate cross-attention layers into recurrent neural networks to facilitate learning the long-term relation dependency and offer an adaptive attention mechanism for varied-length information. Extensive experiments demonstrate that our model exhibits superiority over existing models across KG completion and question-answering tasks.