We present a challenging benchmark for the Open WorLd VISual question answering (OWLViz) task. OWLViz presents concise, unambiguous queries that require integrating multiple capabilities, including visual understanding, web exploration, and specialized tool usage. While humans achieve 69.2% accuracy on these intuitive tasks, even state-of-the-art VLMs struggle, with the best model, Gemini 2.0, achieving only 26.6% accuracy. Current agentic VLMs, which rely on limited vision and vision-language models as tools, perform even worse. This performance gap reveals significant limitations in multimodal systems' ability to select appropriate tools and execute complex reasoning sequences, establishing new directions for advancing practical AI research.

We present FourierNAT, a novel non-autoregressive Transformer (NAT) architecture that employs Fourier-based mixing in the decoder to generate output sequences in parallel. While traditional NAT approaches often face challenges with capturing global dependencies, our method leverages a discrete Fourier transform to mix token embeddings across the entire sequence dimension, coupled with learned frequency-domain gating. This allows the model to efficiently propagate context without explicit autoregressive steps. Empirically, FourierNAT achieves competitive results against leading NAT baselines on standard benchmarks like WMT machine translation and CNN/DailyMail summarization, providing significant speed advantages over autoregressive Transformers. We further demonstrate that learned frequency-domain parameters allow the model to adaptively focus on long-range or short-range dependencies, partially mitigating the well-known coherence gaps in one-pass NAT generation. Overall, FourierNAT highlights the potential of integrating spectral-domain operations to accelerate and improve parallel text generation. This approach can potentially provide great computational and time savings in inference tasks LLMs.

AI industry leaders often use the term ``Jevons' Paradox.'' We explore the significance of this term for artificial intelligence adoption through a time-varying elasticity of substitution framework. We develop a model connecting AI development to labor substitution through four key mechanisms: (1) increased effective computational capacity from both hardware and algorithmic improvements; (2) AI capabilities that rise logarithmically with computation following established neural scaling laws; (3) declining marginal computational costs leading to lower AI prices through competitive pressure; and (4) a resulting increase in the elasticity of substitution between AI and human labor over time. Our time-varying elasticity of substitution (VES) framework, incorporating the G\o rtz identity, yields analytical conditions for market transformation dynamics. This work provides a simple framework to help assess the economic reasoning behind industry claims that AI will increasingly substitute for human labor across diverse economic sectors.

Key variables associated with the direction and intensity of concern include prior experience using a large language model such as ChatGPT, general trust in science, adherence to the precautionary principle versus support for unrestricted innovation, and demographic factors such as gender. By analyzing these relationships, the paper provides valuable insights into the American public's response to AI that are particularly important in the development of policy to regulate or further encourage its development. Key words: artificial intelligence, survey research, public opinion, precautionary principle, ChatGPT Introduction Artificial intelligence is defined as "a machine-based system that can, for a given set of human-defined objectives, make predictions, recommendations, or decisions that impact real or virtual environments" (National Artificial Intelligence Act of 2020, H.R. 6216). According to this definition, AI systems leverage both machine and human inputs to (a) perceive real and virtual environments, (b) transform these perceptions into models through automated analysis, and (c) use these models to generate options for information or action. Currently, AI is employed in a range of applications including mapping technologies, handwriting recognition for mail sorting, spam filtering, language translation, financial trading, and more.

Frontier AI models -- highly capable foundation models at the cutting edge of AI development -- may pose severe risks to public safety, human rights, economic stability, and societal value in the coming years. These risks could arise from deliberate adversarial misuse, system failures, unintended cascading effects, or simultaneous failures across multiple models. In response to such risks, at the AI Seoul Summit in May 2024, 16 global AI industry organizations signed the Frontier AI Safety Commitments, and 27 nations and the EU issued a declaration on their intent to define these thresholds. To fulfill these commitments, organizations must determine and disclose ``thresholds at which severe risks posed by a model or system, unless adequately mitigated, would be deemed intolerable.'' To assist in setting and operationalizing intolerable risk thresholds, we outline key principles and considerations; for example, to aim for ``good, not perfect'' thresholds in the face of limited data on rapidly advancing AI capabilities and consequently evolving risks. We also propose specific threshold recommendations, including some detailed case studies, for a subset of risks across eight risk categories: (1) Chemical, Biological, Radiological, and Nuclear (CBRN) Weapons, (2) Cyber Attacks, (3) Model Autonomy, (4) Persuasion and Manipulation, (5) Deception, (6) Toxicity, (7) Discrimination, and (8) Socioeconomic Disruption. Our goal is to serve as a starting point or supplementary resource for policymakers and industry leaders, encouraging proactive risk management that prioritizes preventing intolerable risks (ex ante) rather than merely mitigating them after they occur (ex post).

Microscopic traffic simulation has become an important tool for autonomous driving training and testing. Although recent data-driven approaches advance realistic behavior generation, their learning still relies primarily on a single real-world dataset, which limits their diversity and thereby hinders downstream algorithm optimization. In this paper, we propose DriveGen, a novel traffic simulation framework with large models for more diverse traffic generation that supports further customized designs. DriveGen consists of two internal stages: the initialization stage uses large language model and retrieval technique to generate map and vehicle assets; the rollout stage outputs trajectories with selected waypoint goals from visual language model and a specific designed diffusion planner. Through this two-staged process, DriveGen fully utilizes large models' high-level cognition and reasoning of driving behavior, obtaining greater diversity beyond datasets while maintaining high realism. To support effective downstream optimization, we additionally develop DriveGen-CS, an automatic corner case generation pipeline that uses failures of the driving algorithm as additional prompt knowledge for large models without the need for retraining or fine-tuning. Experiments show that our generated scenarios and corner cases have a superior performance compared to state-of-the-art baselines. Downstream experiments further verify that the synthesized traffic of DriveGen provides better optimization of the performance of typical driving algorithms, demonstrating the effectiveness of our framework.

Graph neural networks (GNNs) have demonstrated success in modeling relational data primarily under the assumption of homophily. However, many real-world graphs exhibit heterophily, where linked nodes belong to different categories or possess diverse attributes. Additionally, nodes in many domains are associated with textual descriptions, forming heterophilic text-attributed graphs (TAGs). Despite their significance, the study of heterophilic TAGs remains underexplored due to the lack of comprehensive benchmarks. To address this gap, we introduce the Heterophilic Text-attributed Graph Benchmark (HeTGB), a novel benchmark comprising five real-world heterophilic graph datasets from diverse domains, with nodes enriched by extensive textual descriptions. HeTGB enables systematic evaluation of GNNs, pre-trained language models (PLMs) and co-training methods on the node classification task. Through extensive benchmarking experiments, we showcase the utility of text attributes in heterophilic graphs, analyze the challenges posed by heterophilic TAGs and the limitations of existing models, and provide insights into the interplay between graph structures and textual attributes. We have publicly released HeTGB with baseline implementations to facilitate further research in this field.

Large language models (LLMs) have significantly improved their reasoning capabilities; however, they still struggle with complex multi-step mathematical problem-solving due to error propagation, lack of self-correction, and limited adaptability to diverse reasoning styles. Existing methods rely on static fine-tuning or prompt engineering, which fail to generalize across problem complexities, while the scarcity of high-quality preference data further hinders reliable reasoning. We introduce SPHERE, a self-evolving data generation pipeline that enhances reasoning in small language models (SLMs) by iteratively generating, correcting, and diversifying reasoning chains. SPHERE operates in three stages: (i) Self-Generation, where the model autonomously constructs problem-solving steps; (ii) Self-Correction, enabling it to identify and rectify errors; and (iii) Diversity Induction, improving robustness through multiple valid reasoning trajectories. This self-evolution mechanism strengthens mathematical reasoning and enhances model reliability. Evaluations on MATH 500, GSM8K, AIME, AMC, and Olympiad show that SPHERE-trained models achieve significant gains over their base versions and match/surpass GPT-4o on certain benchmarks. Our findings demonstrate that self-evolving models can close the reasoning gap between SLMs and state-of-the-art LLMs, making mathematical AI more reliable, scalable, and efficient.

Universal multimodal embedding models play a critical role in tasks such as interleaved image-text retrieval, multimodal RAG, and multimodal clustering. However, our empirical results indicate that existing LMM-based embedding models trained with the standard InfoNCE loss exhibit a high degree of overlap in similarity distribution between positive and negative pairs, making it challenging to distinguish hard negative pairs effectively. To deal with this issue, we propose a simple yet effective framework that dynamically improves the embedding model's representation learning for negative pairs based on their discriminative difficulty. Within this framework, we train a series of models, named LLaVE, and evaluate them on the MMEB benchmark, which covers 4 meta-tasks and 36 datasets. Experimental results show that LLaVE establishes stronger baselines that achieve state-of-the-art (SOTA) performance while demonstrating strong scalability and efficiency. Specifically, LLaVE-2B surpasses the previous SOTA 7B models, while LLaVE-7B achieves a further performance improvement of 6.2 points. Although LLaVE is trained on image-text data, it can generalize to text-video retrieval tasks in a zero-shot manner and achieve strong performance, demonstrating its remarkable potential for transfer to other embedding tasks.

Social bots have become widely known by users of social platforms. To prevent social bots from spreading harmful speech, many novel bot detections are proposed. However, with the evolution of social bots, detection methods struggle to give high-confidence answers for samples. This motivates us to quantify the uncertainty of the outputs, informing the confidence of the results. Therefore, we propose an uncertainty-aware bot detection method to inform the confidence and use the uncertainty score to pick a high-confidence decision from multiple views of a social network under different environments. Specifically, our proposed BotUmc uses LLM to extract information from tweets. Then, we construct a graph based on the extracted information, the original user information, and the user relationship and generate multiple views of the graph by causal interference. Lastly, an uncertainty loss is used to force the model to quantify the uncertainty of results and select the result with low uncertainty in one view as the final decision. Extensive experiments show the superiority of our method.