Recent embodied navigation approaches leveraging Vision-Language Models (VLMs) demonstrate strong generalization in versatile Vision-Language Navigation (VLN). However, reliable path planning in complex environments remains challenging due to insufficient spatial awareness. In this work, we introduce SPAN-Nav, an end-to-end foundation model designed to infuse embodied navigation with universal 3D spatial awareness using RGB video streams. SPAN-Nav extracts spatial priors across diverse scenes through an occupancy prediction task on extensive indoor and outdoor environments. To mitigate the computational burden, we introduce a compact representation for spatial priors, finding that a single token is sufficient to encapsulate the coarse-grained cues essential for navigation tasks. Furthermore, inspired by the Chain-of-Thought (CoT) mechanism, SPAN-Nav utilizes this single spatial token to explicitly inject spatial cues into action reasoning through an end-to end framework. Leveraging multi-task co-training, SPAN-Nav captures task-adaptive cues from generalized spatial priors, enabling robust spatial awareness to generalize even to the task lacking explicit spatial supervision. To support comprehensive spatial learning, we present a massive dataset of 4.2 million occupancy annotations that covers both indoor and outdoor scenes across multi-type navigation tasks. SPAN-Nav achieves state-of-the-art performance across three benchmarks spanning diverse scenarios and varied navigation tasks. Finally, real-world experiments validate the robust generalization and practical reliability of our approach across complex physical scenarios.

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, particularly for long-tail domains such as literary works. However, the critical step of document segmentation in RAG remains largely underexplored. Existing strategies are typically semantically blind and overlook the complicated narrative structures of literary works, often resulting in fragmented plots and unclear references that severely hinder retrieval and generation performance. To address this, we propose LitSeg, a novel narrative-theory-guided segmentation framework. By employing multi-stage prompting, LitSeg explicitly extracts valid events, untangles narrative threads, clarifies narrative structures, and locates turning points to inform segmentation. To alleviate the computational overhead of multi-stage inference with large-scale models, we further introduce LitSeg-Lite, a lightweight single-pass chunker fine-tuned on LitSeg-generated data via a two-stage training strategy, distilling the complex process into a single inference pass. Extensive experiments demonstrate that with structurally independent text chunks, our methods significantly improve retrieval accuracy and context relevance over baselines, ultimately enhancing downstream QA performance, while ablation studies validate the efficacy of narratological guidance and data distillation.

Inline notes and collected commentaries are important forms of scholarly communication that evolved within the Confucian exegetical tradition, yet have received little computational attention. Drawing on traditional Chinese exegetics and philology, this paper formulates collected commentary compilation as an NLP task and proposes a computational framework that preserves the contextual dependency of inline notes while enabling their automatic compilation and exegetical knowledge organization. It combines two-step prompt chaining for identifying the associated main-text segments and exegetical functions of annotations with cross-source mention clustering for integrating commentary across editions, achieving a CoNLL F1 score above 97

LLM-based research agents have advanced rapidly in science and engineering, where research is organized around executable experiments, code, and quantitative signals. Humanities scholarship, however, requires a different mode of reasoning: interpretive, evidence-grounded argument over primary sources, where scholarly value depends on faithful quotation, verifiable provenance, and close reading. Existing research agents remain largely optimized for execution and retrieval, not evidence-grounded interpretive reasoning. To address this gap, we introduce SPIRE (Scholarly-Primitives-Inspired Research Engine), a multi-agent framework for evidence-grounded humanities scholarship. Drawing on Scholarly Primitives theory, SPIRE casts recurring humanities operations as cooperating agent roles (source discovery, evidence annotation, comparison, provenance checking, sampling, citation binding, and argumentative synthesis) over a multi-scale close-reading substrate of passages, intra-context graph communities, and cross-context semantic clusters. On a peer-reviewed-paper benchmark over classical Chinese and Greco-Roman Latin scholarship, SPIRE recovers cited primary-source evidence more reliably than Naive LLM, Text RAG, and GraphRAG, and receives higher blind-judge scores on answer accuracy, depth, coverage, and evidence quality. Ablations show that both the scholarly-operation agents and close-reading retrieval contribute to evidence-grounded essays. Code, data catalogues, and reproduction scripts are released at https://github.com/YatingPan/SPIRE.

Humor, as both a creative human activity and a social binding mechanism, has long posed a major challenge for AI generation. Although producing humor requires complex cognitive reasoning and social understanding, theories of humor suggest that it follows learnable patterns and structures, making it theoretically possible for generative models to acquire them implicitly. In recent years, multimodal humor has become a prevalent form of online communication, especially among Gen Z, highlighting the need for AI systems capable of integrating visual understanding with humorous language generation. However, existing data-driven approaches lack explicit modeling or theoretical grounding of humor, often producing literal descriptions that fail to capture its underlying cognitive mechanisms, resulting in the generated image descriptions that are fluent but lack genuine humor or cognitive depth. To address this limitation, we propose HUMORCHAIN (HUmor-guided Multi-step Orchestrated Reasoning Chain for Image Captioning), a theory-guided multi-stage reasoning framework. It integrates visual semantic parsing, humor- and psychology-based reasoning, and a fine-tuned discriminator for humor evaluation, forming an interpretable and controllable cognitive reasoning chain. To the best of our knowledge, this is the first work to explicitly embed cognitive structures from humor theories into multimodal humor generation, enabling a structured reasoning process from visual understanding to humor creation. Experiments on Meme-Image-No-Text, Oogiri-GO, and OxfordTVG-HIC datasets show that HUMORCHAIN outperforms state-of-the-art baselines in human humor preference, Elo/BT scores, and semantic diversity, demonstrating that theory-driven structured reasoning enables large language models to generate humor aligned with human perception.

Vision-Language-Action (VLA) policies have emerged as a versatile paradigm for generalist robotic manipulation. However, precise object placement under compositional language remains challenging for end-to-end VLA policies. Slot-level placement requires reliable slot grounding and centimeter-level geometric precision. To this end, we propose AnySlot, a framework that reduces compositional complexity by introducing an explicit spatial visual goal between language grounding and control. AnySlot converts language into a visual goal by rendering a spatial marker at the intended slot, then executes this goal with a goal-conditioned VLA policy. This hierarchical design decouples high-level slot selection from low-level execution, improving semantic accuracy and spatial robustness. Furthermore, recognizing the lack of benchmarks for such precision-demanding tasks, we introduce SlotBench, a structured simulation benchmark with nine task categories for evaluating spatial reasoning in slot-level placement. Extensive experiments show that AnySlot significantly outperforms flat VLA baselines and modular grounding methods in zero-shot slot-level placement.

Human personality theories characterize traits not as isolated attributes captured by a single score, but as stable individual tendencies expressed through the interplay among persons, situations, and behaviors. Existing studies of personality-related behavior in LLMs have primarily focused on outputs elicited under personality conditioning, characterizing observable trait-related expressions while lacking mechanistic evidence for the existence of internal personality-related representations, their cross-situational expression, and how these representations shape specific behaviors. Building on Funder's personality triad framework, we adapt its three components for LLM analysis: Person as personality-related internal representations, Situation as contexts that afford trait-relevant responses, and Behavior as response patterns on broader social tasks. We introduce a framework for discovering, controlling, and validating trait-like representations in LLMs. First, using contrastive behavior pairs grounded in shared situations, we identify sparse internal features associated with opposing poles of personality traits through SAE decomposition. We validate their trait relevance through effects on behavior to situation, token-level activation patterns, and robustness to paraphrasing. Second, feature-level interventions induce bidirectional trait-related shifts across a separate, diverse set of situations while preserving response validity, demonstrating consistent expression across contexts. Third, applying the same interventions to social intelligence tasks reveals behavioral changes with benefit-tradeoff patterns consistent with findings from human personality research, providing behavioral-level validation beyond personality scores. Our findings provide evidence that LLMs contain controllable trait-like representations linking internal states, situational expression, and behavioral outcomes.

Recent work in Mechanistic Interpretability (MI) has enabled the identification and intervention of internal features in Large Language Models (LLMs). However, a persistent challenge lies in linking such internal features to the reliable control of complex, behavior-level semantic attributes in language generation. In this paper, we propose a Sparse Autoencoder-based framework for retrieving and steering semantically interpretable internal features associated with high-level linguistic behaviors. Our method employs a contrastive feature retrieval pipeline based on controlled semantic oppositions, combing statistical activation analysis and generation-based validation to distill monosemantic functional features from sparse activation spaces. Using the Big Five personality traits as a case study, we demonstrate that our method enables precise, bidirectional steering of model behavior while maintaining superior stability and performance compared to existing activation steering methods like Contrastive Activation Addition (CAA). We further identify an empirical effect, which we term Functional Faithfulness, whereby intervening on a specific internal feature induces coherent and predictable shifts across multiple linguistic dimensions aligned with the target semantic attribute. Our findings suggest that LLMs internalize deeply integrated representations of high-order concepts, and provide a novel, robust mechanistic path for the regulation of complex AI behaviors.

The rapid evolution of communication technologies and the Internet of Things has enhanced system interconnectivity, allowing localized disruptions to cascade into failures. Many such breakdowns are not purely accidental but emerge from intentional countermeasures that influence one another during disruptions. Capturing this interplay requires a framework describing how strategic behaviors coevolve and collectively determine the resilience of complex networks. Here we develop a game-theoretic framework that captures the coevolution of attacks and defenses through repeated interactions on a networked system. Attackers and defenders update strategies based on past outcomes, generating adaptive dynamics that link network structure with strategic behavior. We examine strategic unilateral control of payoffs and uncover a heterogeneity-dependent asymmetry: attackers dominate on heterogeneous networks, whereas defenders prevail on homogeneous ones. Remarkably, optimal attacker strategies converge on the network's percolation threshold, as confirmed across synthetic and empirical networks. Our results reveal how topology governs strategic coevolution and emergent resilience.

Current Transformer models with channel independence (CI) have made tremendous achievements in time series data analysis. However, the CI methods suffer from short-term fluctuations with intra-channel noise and long-term trend extraction. The fixed receptive field of CI models struggles with capturing multi-scale temporal features within each channel. This paper proposes CSFformer, a cross-scale fusion Transformer for multivariate time series. We first present a Channel-Independent Masking (CIM) module to rectify the feature representation of anomalies and noise in each channel. Then we design a Multi-Scale Pyramid Fusion (MSPF) module, which extracts fluctuation and tread features across various scales. Furthermore, the Multi-Scale Attention Fusion (MSAF) module is introduced for in-depth analysis of interactions between different scales, which significantly contributes to capturing a broader spectrum of complex temporal patterns. We conduct experiments on 7 real-world public datasets. The results show that CSFformer achieves state-of-the-art performance on all datasets, especially in scenarios with obvious fluctuations and trends such as Traffic and Electricity. The source code is available at: https://github.com/damonwan1/CSIformer.

This work contributes to the digital humanities approach for studying premodern Chinese history and culture by creating a large-scale dataset annotated with named entities and relations. Through careful annotation guidelines and labeling of over 200,000 characters, we developed a dataset containing 30,000 named entities across six types and 7,000 relations spanning twenty categories. Experiments on named entity recognition (NER) using pre-trained language models and large language models on this dataset achieved an initial performance of NER (91.32 percent F1). In addition, relationship extraction (RE) on the pretrained language model achieves an 85.32 percent F1 score. While there is still room for improvement, our annotated dataset and models provide a useful starting point for extracting semantic information from premodern Chinese texts. It represents an effort to connect history and technology, increasing accessibility and preservation of premodern Chinese cultural treasures. Furthermore, our dataset can facilitate downstream tasks like culture analysis, knowledge graph construction, and computational understanding of premodern Chinese. Overall, this research represents a significant step toward digitally exploring premodern Chinese documents, providing a pathway for future work on knowledge organization and computational analysis of this valuable cultural legacy. Our code and data are available at: https://github.com/tangxuemei1995/AnChineseNERE

Idiopathic pulmonary fibrosis (IPF) is a severe lung disease occurring throughout the world; however, few clinical therapies are available for treating this disorder. Overactivated fibroblasts drive abnormal fibrosis accumulation to maintain dynamic balance between inflammation and extracellular matrix (ECM) stiffness. Given pulmonary cell can regenerate, the lung may possess self-repairing abilities if fibrosis is removed via clearance of overactivated fibroblasts. The aim of this study was to evaluate the therapeutic activity of transient antifibrotic chimeric antigen receptor (CAR) T cells (generated via a novelly-designed lipid nanoparticle-messenger RNA (LNP-mRNA) system) and explore the regeneration mechanisms of lung in a male mouse model of bleomycin-induced pulmonary fibrosis. Here we found that fibrosis-induced ECM stiffening impaired alveolar epithelial cell compensation. The proposed LNP-mRNA therapy eliminated overactivated fibroblasts to rescue pulmonary fibrosis. The restored ECM environment regulated the cellular profile. The elevated plasticity of AT2 and Pclaf+ cells increased AT1 cell population via polarization. Apoe+ macrophages and increased numbers of effector T cells were shown to reestablish pulmonary immunity. Hence, LNP-mRNA treatment for fibrosis can restore pulmonary structure and function to similar degrees to those of a healthy lung. This therapy is a potential treatment for IPF patients.

Sequence labeling models often benefit from incorporating external knowledge. However, this practice introduces data heterogeneity and complicates the model with additional modules, leading to increased expenses for training a high-performing model. To address this challenge, we propose a two-stage curriculum learning (TCL) framework specifically designed for sequence labeling tasks. The TCL framework enhances training by gradually introducing data instances from easy to hard, aiming to improve both performance and training speed. Furthermore, we explore different metrics for assessing the difficulty levels of sequence labeling tasks. Through extensive experimentation on six Chinese word segmentation (CWS) and Part-of-speech tagging (POS) datasets, we demonstrate the effectiveness of our model in enhancing the performance of sequence labeling models. Additionally, our analysis indicates that TCL accelerates training and alleviates the slow training problem associated with complex models.

Recent advances in unsupervised sentence representation learning have relied heavily on contrastive objectives over pre-trained language models (PLMs). However, the quality of learned representations is often limited by shallow layers and restricted model capacity, especially in the absence of large-scale supervision. To address this, we propose SageRep, a novel layer-adaptive self-knowledge distillation framework that improves representation quality without requiring external teacher models or multi-stage training. Motivated by the observation that deeper layers in PLMs encode richer semantic signals, SageRep enables shallower layers to distill knowledge dynamically from deeper ones, guided by sentence-level similarity distributions. Unlike prior static self-knowledge distillation methods, our approach adaptively selects the most informative teacher layers per instance, promoting more effective intra-model knowledge transfer. Additionally, we introduce a contrastive regularization strategy using inter-layer negatives to mitigate representation over-smoothing. Extensive evaluations on standard semantic textual similarity benchmarks demonstrate that SageRep achieves superior performance over previous unsupervised methods, with minimal additional training cost.

Object detection in autonomous driving suffers from motion blur and saturation under fast motion and extreme lighting. Spike cameras, offer microsecond latency and ultra high dynamic range for object detection by using per pixel asynchronous integrate and fire. However, their sparse, discrete output cannot be processed by standard image-based detectors, posing a critical challenge for end to end spike stream detection. We propose EASD, an end to end spike camera detector with a dual branch design: a Temporal Based Texture plus Feature Fusion branch for global cross slice semantics, and an Entropy Selective Attention branch for object centric details. To close the data gap, we introduce DSEC Spike, the first driving oriented simulated spike detection benchmark.

Exploring dominant strategies in iterated games holds theoretical and practical significance across diverse domains. Previous studies, through mathematical analysis of limited cases, have unveiled classic strategies such as tit-for-tat, generous-tit-for-tat, win-stay-lose-shift, and zero-determinant strategies. While these strategies offer valuable insights into human decision-making, they represent only a small subset of possible strategies, constrained by limited mathematical and computational tools available to explore larger strategy spaces. To bridge this gap, we propose an approach using multi-agent reinforcement learning to delve into complex decision-making processes that go beyond human intuition. Our approach has led to the discovery of a strategy that we call memory-two bilateral reciprocity strategy. Memory-two bilateral reciprocity strategy consistently outperforms a wide range of strategies in pairwise interactions while achieving high payoffs. When introduced into an evolving population with diverse strategies, memory-two bilateral reciprocity strategy demonstrates dominance and fosters higher levels of cooperation and social welfare in both homogeneous and heterogeneous structures, as well as across various game types. This high performance is verified by simulations and mathematical analysis. Our work highlights the potential of multi-agent reinforcement learning in uncovering dominant strategies in iterated and evolutionary games.

Network resilience generally refers to the ability of a network to maintain normal operation in the face of various failures and disturbances, which has many applications in social, financial, communication, and transportation systems. There is a large amount of research on designing strategies to disrupt the network from the perspective of attack and to protect the network from the defense standpoint. However, attacks and defenses in the real world often occur simultaneously and repeatedly. The study of network resilience should not be limited to the structural characteristics but should consider the functional characteristics of the network. We develop a general game-theoretic framework to study attack-defense repeated interactions on complex net-works and examine the possibility of achieving precise control over the payoffs of attackers and defenders. Our findings reveal that neither the attacker nor the defender can unilaterally control their own payoff, the opponent's payoff, or any linear combination thereof. This work highlights the limitations of unilateral payoff control in real-world interaction scenarios, a consideration often overlooked in numerous theoretical studies.

Cooperation is key to prosperity in human societies. Population structure is well understood as a catalyst for cooperation, where research has focused on pairwise interactions. But cooperative behaviors are not simply dyadic, and they often involve coordinated behavior in larger groups. Here we develop a framework to study the evolution of behavioral strategies in higher-order population structures, which include pairwise and multi-way interactions. We provide an analytical treatment of when cooperation will be favored by higher-order interactions, accounting for arbitrary spatial heterogeneity and nonlinear rewards for cooperation in larger groups. Our results indicate that higher-order interactions can act to promote the evolution of cooperation across a broad range of networks, in public goods games. Higher-order interactions consistently provide an advantage for cooperation when interaction hyper-networks feature multiple conjoined communities. Our analysis provides a systematic account of how higher-order interactions modulate the evolution of prosocial traits. Cooperation is not merely a dyadic phenomenon, it also includes multi-way social interactions. A mathematical framework is developed to study how the structure of higher-order interactions influences cooperative behavior.

Exploring dominant strategies in iterated games holds profound theoretical and practical significance across diverse domains. Previous studies, through mathematical analysis of limited cases, have unveiled classic strategies such as tit-for-tat, generous-tit-for-tat, win-stay-lose-shift, and zero-determinant strategies. While these strategies offer valuable insights into human decision-making, they represent only a small subset of possible strategies, constrained by limited mathematical and computational tools available to explore larger strategy spaces. To bridge this gap, we propose a novel approach using multi-agent reinforcement learning to delve into complex decision-making processes that go beyond human intuition. Our approach has led to the discovery of a new strategy, i.e., memory-two bilateral reciprocity (MTBR) strategy. MTBR consistently outperforms a wide range of strategies in pairwise interactions while achieving high payoffs. When introduced into an evolving population with diverse strategies, MTBR demonstrates dominance and fosters higher levels of cooperation and social welfare in both homogeneous and heterogeneous structures, as well as across various game types. This high performance is verified by simulations and mathematical analysis. Our work highlights the potential of multi-agent reinforcement learning in uncovering dominant strategies in complex environments, offering a new perspective on strategy exploration in iterated games.

In this study, we propose a new evaluation scheme to assess the strengths and limitations of collocation extraction measures and explore type-sensitive methods for extracting collocations. We introduced the pooling strategy widely used in Information Retrieval and automated the evaluation process using online dictionaries. Sixteen well-known metrics are evaluated based on their effectiveness and then distributional and linguistic compared. The results show that Group A methods (e.g. z-score, Dice, PMI) are more effective in extracting low-frequency collocations with relatively small extraction scales. In contrast, Group B methods (e.g. t-test, LMI, LLR) perform better at finding high-frequency collocations, most of which outperform Group A methods as the extraction scale increases. Moreover, Group A prefers NN collocations, while Group B identifies collocations with a wide range of syntactic structures. This study provides suggestions for studies to identify hybrid extraction methods as well as for language educators and dictionary compilers.