Understanding what constitutes safety in AI-generated content is complex. While developers often rely on predefined taxonomies, real-world safety judgments also involve personal, social, and cultural perceptions of harm. This paper examines how annotators evaluate the safety of AI-generated images, focusing on the qualitative reasoning behind their judgments. Analyzing 5,372 open-ended comments, we find that annotators consistently invoke moral, emotional, and contextual reasoning that extends beyond structured safety categories. Many reflect on potential harm to others more than to themselves, grounding their judgments in lived experience, collective risk, and sociocultural awareness. Beyond individual perceptions, we also find that the structure of the task itself -- including annotation guidelines -- shapes how annotators interpret and express harm. Guidelines influence not only which images are flagged, but also the moral judgment behind the justifications. Annotators frequently cite factors such as image quality, visual distortion, and mismatches between prompt and output as contributing to perceived harm dimensions, which are often overlooked in standard evaluation frameworks. Our findings reveal that existing safety pipelines miss critical forms of reasoning that annotators bring to the task. We argue for evaluation designs that scaffold moral reflection, differentiate types of harm, and make space for subjective, context-sensitive interpretations of AI-generated content.

Human intelligence exhibits a remarkable capacity for rapid adaptation and effective problem-solving in novel and unfamiliar contexts. We argue that this profound adaptability is fundamentally linked to the efficient construction and refinement of internal representations of the environment, commonly referred to as world models, and we refer to this adaptation mechanism as world model induction . However, current understanding and evaluation of world models in artificial intelligence (AI) remains narrow, often focusing on static representations learned from training on massive corpora of data, instead of the efficiency and efficacy in learning these representations through interaction and exploration within a novel environment. In this Perspective, we provide a view of world model induction drawing on decades of research in cognitive science on how humans learn and adapt so efficiently; we then call for a new evaluation framework for assessing adaptive world models in AI. Concretely, we propose a new benchmarking paradigm based on suites of carefully designed games with genuine, deep and continually refreshing novelty in the underlying game structures -- we refer to this class of games as novel games . We detail key desiderata for constructing these games and propose appropriate metrics to explicitly challenge and evaluate the agent's ability for rapid world model induction. We hope that this new evaluation framework will inspire future evaluation efforts on world models in AI and provide a crucial step towards developing AI systems capable of human-like rapid adaptation and robust generalization -- a critical component of artificial general intelligence.

The diagnosis of oral diseases presents a problematic clinical challenge, characterized by a wide spectrum of pathologies with overlapping symptomatology. To address this, we developed Clinical Semantic Intelligence (CSI), a novel artificial intelligence framework that diagnoses 118 different oral diseases by computationally modeling the cognitive processes of an expert clinician. Our core hypothesis is that moving beyond simple pattern matching to emulate expert reasoning is critical to building clinically useful diagnostic aids. CSI's architecture integrates a fine-tuned multimodal CLIP model with a specialized ChatGLM-6B language model. This system executes a Hierarchical Diagnostic Reasoning Tree (HDRT), a structured framework that distills the systematic, multi-step logic of differential diagnosis. The framework operates in two modes: a Fast Mode for rapid screening and a Standard Mode that leverages the full HDRT for an interactive and in-depth diagnostic workup. To train and validate our system, we curated a primary dataset of 4,310 images, supplemented by an external hold-out set of 176 images for final validation. A clinically-informed augmentation strategy expanded our training data to over 30,000 image-text pairs. On a 431-image internal test set, CSI's Fast Mode achieved an accuracy of 73.4%, which increased to 89.5% with the HDRT-driven Standard Mode. The performance gain is directly attributable to the hierarchical reasoning process. Herein, we detail the architectural philosophy, development, and rigorous evaluation of the CSI framework.

Reasoning distillation has emerged as an effective approach to enhance the reasoning capabilities of smaller language models. However, the impact of large-scale reasoning distillation on other critical abilities, particularly in-context retrieval and reasoning, remains unexplored. This gap in understanding is particularly significant given the increasing importance of Retrieval-Augmented Generation (RAG) systems, where efficient acquisition and utilization of contextual information are paramount for generating reliable responses. Motivated by the need to understand how the extended long-CoT process influences long-context comprehension, we conduct a comprehensive investigation using a series of open-source models distilled from Deepseek-R1, renowned for its exceptional reasoning capabilities. Our study focuses on evaluating these models' performance in extracting and integrating relevant information from extended contexts through multi-document question and answering tasks. Through rigorous experimentation, we demonstrate that distilled reasoning patterns significantly improve long-context understanding. Our analysis reveals that distillation fosters greater long-context awareness by promoting more detailed and explicit reasoning processes during context analysis and information parsing. This advancement effectively mitigates the persistent "lost in the middle" issue that has hindered long-context models.

Although existing models can interact with humans and provide satisfactory responses, they lack the ability to act autonomously or engage in independent reasoning. Furthermore, input data in these models is typically provided as explicit queries, even when some sensory data is already acquired. In addition, AI agents, which are computational entities designed to perform tasks and make decisions autonomously based on their programming, data inputs, and learned knowledge, have shown significant progress. However, they struggle with integrating knowledge across multiple domains, unlike humans. Mental imagery plays a fundamental role in the brain's thinking process, which involves performing tasks based on internal multisensory data, planned actions, needs, and reasoning capabilities. In this paper, we investigate how to integrate mental imagery into a machine thinking framework and how this could be beneficial in initiating the thinking process. Our proposed machine thinking framework integrates a Cognitive thinking unit supported by three auxiliary units: the Input Data Unit, the Needs Unit, and the Mental Imagery Unit. Within this framework, data is represented as natural language sentences or drawn sketches, serving both informative and decision-making purposes. We conducted validation tests for this framework, and the results are presented and discussed.

Synthesising safe controllers from visual data typically requires extensive supervised labelling of safety-critical data, which is often impractical in real-world settings. Recent advances in world models enable reliable prediction in latent spaces, opening new avenues for scalable and data-efficient safe control. In this work, we introduce a semi-supervised framework that leverages control barrier certificates (CBCs) learned in the latent space of a world model to synthesise safe visuomotor policies. Our approach jointly learns a neural barrier function and a safe controller using limited labelled data, while exploiting the predictive power of modern vision transformers for latent dynamics modelling.

Transformers evaluated in a single, fixed-depth pass are provably limited in expressive power to the constant-depth circuit class TC0. Running a Transformer autoregressively removes that ceiling -- first in next-token prediction and, more recently, in chain-of-thought reasoning. Both regimes rely on feedback loops that decode internal states into tokens only to re-encode them in subsequent steps. While this "thinking aloud" mirrors human reasoning, biological brains iterate without externalising intermediate states as language. To boost the expressive power of encoder Transformers without resorting to token-level autoregression, we introduce the SELF-Transformer: an encoder layer that iteratively refines its own attention weights to a fixed point. Instead of producing -- in one pass -- the alignment matrix that remixes the input sequence, the SELF-Transformer iteratively updates that matrix internally, scaling test-time computation with input difficulty. This adaptivity yields up to 20\% accuracy gains on encoder-style benchmarks without increasing parameter count, demonstrating that input-adaptive alignment at test time offers substantial benefits for only a modest extra compute budget. Self-Transformers thus recover much of the expressive power of iterative reasoning while preserving the simplicity of pure encoder architectures.

Personalizing AI systems requires understanding not just what users prefer, but the reasons that underlie those preferences - yet current preference models typically treat human judgment as a black box. We introduce PrefPalette, a framework that decomposes preferences into attribute dimensions and tailors its preference prediction to distinct social community values in a human-interpretable manner. PrefPalette operationalizes a cognitive science principle known as multi-attribute decision making in two ways: (1) a scalable counterfactual attribute synthesis step that involves generating synthetic training data to isolate for individual attribute effects (e.g., formality, humor, cultural values), and (2) attention-based preference modeling that learns how different social communities dynamically weight these attributes. This approach moves beyond aggregate preference modeling to capture the diverse evaluation frameworks that drive human judgment. When evaluated on 45 social communities from the online platform Reddit, PrefPalette outperforms GPT-4o by 46.6% in average prediction accuracy. Beyond raw predictive improvements, PrefPalette also shed light on intuitive, community-specific profiles: scholarly communities prioritize verbosity and stimulation, conflict-oriented communities value sarcasm and directness, and support-based communities emphasize empathy. By modeling the attribute-mediated structure of human judgment, PrefPalette delivers both superior preference modeling and transparent, interpretable insights, and serves as a first step toward more trustworthy, value-aware personalized applications.

Graph Retrieval-Augmented Generation has emerged as a powerful paradigm for grounding large language models with external structured knowledge. However, existing Graph RAG methods struggle with temporal reasoning, due to their inability to model the evolving structure and order of real-world events. In this work, we introduce DyG-RAG, a novel event-centric dynamic graph retrieval-augmented generation framework designed to capture and reason over temporal knowledge embedded in unstructured text. To eliminate temporal ambiguity in traditional retrieval units, DyG-RAG proposes Dynamic Event Units (DEUs) that explicitly encode both semantic content and precise temporal anchors, enabling accurate and interpretable time-aware retrieval. To capture temporal and causal dependencies across events, DyG-RAG constructs an event graph by linking DEUs that share entities and occur close in time, supporting efficient and meaningful multi-hop reasoning. To ensure temporally consistent generation, DyG-RAG introduces an event timeline retrieval pipeline that retrieves event sequences via time-aware traversal, and proposes a Time Chain-of-Thought strategy for temporally grounded answer generation. This unified pipeline enables DyG-RAG to retrieve coherent, temporally ordered event sequences and to answer complex, time-sensitive queries that standard RAG systems cannot resolve. Extensive experiments on temporal QA benchmarks demonstrate that DyG-RAG significantly improves the accuracy and recall of three typical types of temporal reasoning questions, paving the way for more faithful and temporal-aware generation. DyG-RAG is available at https://github .com/RingBDStack/DyG-RAG.

Traditional agent-based urban mobility simulations often rely on rigid rule-based systems that struggle to capture the complexity, adaptability, and behavioral diversity inherent in human travel decision making. Recent advancements in large language models and AI agent technologies present new opportunities to develop agents with enhanced reasoning capabilities, persistent memory, and adaptive learning. We introduce GATSim (Generative-Agent Transport Simulation), a novel framework that leverages these advancements to simulate urban mobility using generative agents with rich, human-like behaviors. Unlike conventional approaches, GATSim agents are characterized by diverse socioeconomic profiles, individual lifestyles, and evolving preferences shaped through psychologically informed memory systems, tool usage, and lifelong learning. The main contributions of this work are: (1) a comprehensive architecture that integrates an urban mobility foundation model with agent cognitive systems and a transport simulation environment; (2) a hierarchical memory designed for efficient retrieval of contextually relevant information, incorporating spatial and temporal associations, keyword matching, and semantic relevance; (3) innovative planning and reactive mechanisms for modeling adaptive mobility behaviors which integrate a multi-scale reflection process to transform specific travel experiences into generalized behavioral insights. We implement a prototype system and conduct systematic validation, demonstrating that generative agents produce believable and coherent travel behaviors. Experimental results indicate that generative agents perform at least as well as human annotators with 92\% posterior probability, while naturally producing realistic macroscopic traffic patterns. The code for the prototype implementation is publicly available at https://github.com/qiliuchn/gatsim.