Language model deployments in consumer-facing applications introduce numerous risks. While existing research on harms and hazards of such applications follows top-down approaches derived from regulatory frameworks and theoretical analyses, empirical evidence of real-world failure modes remains underexplored. In this work, we introduce RealHarm, a dataset of annotated problematic interactions with AI agents built from a systematic review of publicly reported incidents. Analyzing harms, causes, and hazards specifically from the deployer's perspective, we find that reputational damage constitutes the predominant organizational harm, while misinformation emerges as the most common hazard category. We empirically evaluate state-of-the-art guardrails and content moderation systems to probe whether such systems would have prevented the incidents, revealing a significant gap in the protection of AI applications.

With the shift from traditional to digital media, the online landscape now hosts not only reliable news articles but also a significant amount of unreliable content. Digital media has faster reachability by significantly influencing public opinion and advancing political agendas. While newspaper readers may be familiar with their preferred outlets political leanings or credibility, determining unreliable news articles is much more challenging. The credibility of many online sources is often opaque, with AI generated content being easily disseminated at minimal cost. Unreliable news articles, particularly those that followed the Russian invasion of Ukraine in 2022, closely mimic the topics and writing styles of credible sources, making them difficult to distinguish. To address this, we introduce SemCAFE, a system designed to detect news reliability by incorporating entity relatedness into its assessment. SemCAFE employs standard Natural Language Processing techniques, such as boilerplate removal and tokenization, alongside entity level semantic analysis using the YAGO knowledge base. By creating a semantic fingerprint for each news article, SemCAFE could assess the credibility of 46,020 reliable and 3,407 unreliable articles on the 2022 Russian invasion of Ukraine. Our approach improved the macro F1 score by 12% over state of the art methods. The sample data and code are available on GitHub

Pretraining data curation is a cornerstone in Large Language Model (LLM) development, leading to growing research on quality filtering of large web corpora. From statistical quality flags to LLM-based labelling systems, datasets are divided into categories, frequently reducing to a binary: those passing the filters are deemed as valuable examples, others are discarded as useless or detrimental. However, a more detailed understanding of the contribution of different kinds of texts to model performance is still largely lacking. In this article, we present the first study utilising registers or genres - a widely used standard in corpus linguistics to model linguistic variation - to curate pretraining datasets and investigate the effect of register on the performance of LLMs. We train small generative models with register classified data and evaluate them using standard benchmarks, and show that the register of pretraining data substantially affects model performance. We uncover surprising relationships between the pretraining material and the resulting models: using the News register results in subpar performance, and on the contrary, including the Opinion class, covering texts such as reviews and opinion blogs, is highly beneficial. While a model trained on the entire unfiltered dataset outperforms those trained on datasets limited to a single register, combining well-performing registers like How-to-Instructions, Informational Description, and Opinion leads to major improvements. Furthermore, analysis of individual benchmark results reveals key differences in the strengths and drawbacks of specific register classes as pretraining data. These findings show that register is an important explainer of model variation and can facilitate more deliberate future data selection practices.

Recent advances in large multimodal models have leveraged image-based tools with reinforcement learning to tackle visual problems. However, existing open-source approaches often exhibit monotonous reasoning patterns and allow only a limited number of interaction turns, making them inadequate for difficult tasks that require trial-and-error exploration. In this work, we address this limitation by scaling up tool-based interactions and introduce Mini-o3, a system that executes deep, multi-turn reasoning -- spanning tens of steps -- and achieves state-of-the-art performance on challenging visual search tasks. Our recipe for reproducing OpenAI o3-style behaviors comprises three key components. First, we construct the Visual Probe Dataset, a collection of thousands of challenging visual search problems designed for exploratory reasoning. Second, we develop an iterative data collection pipeline to obtain cold-start trajectories that exhibit diverse reasoning patterns, including depth-first search, trial-and-error, and goal maintenance. Third, we propose an over-turn masking strategy that prevents penalization of over-turn responses (those that hit the maximum number of turns) during reinforcement learning, thereby balancing training-time efficiency with test-time scalability. Despite training with an upper bound of only six interaction turns, our model generates trajectories that naturally scale to tens of turns at inference time, with accuracy improving as the number of turns increases. Extensive experiments demonstrate that Mini-o3 produces rich reasoning patterns and deep thinking paths, effectively solving challenging visual search problems.

We introduce SimpleQA Verified, a 1,000-prompt benchmark for evaluating Large Language Model (LLM) short-form factuality based on OpenAI's SimpleQA. It addresses critical limitations in OpenAI's benchmark, including noisy and incorrect labels, topical biases, and question redundancy. SimpleQA Verified was created through a rigorous multi-stage filtering process involving de-duplication, topic balancing, and source reconciliation to produce a more reliable and challenging evaluation set, alongside improvements in the autorater prompt. On this new benchmark, Gemini 2.5 Pro achieves a state-of-the-art F1-score of 55.6, outperforming other frontier models, including GPT-5. This work provides the research community with a higher-fidelity tool to track genuine progress in parametric model factuality and to mitigate hallucinations. The benchmark dataset, evaluation code, and leaderboard are available at: https://www.kaggle.com/benchmarks/deepmind/simpleqa-verified.

Visual reasoning over structured data such as tables is a critical capability for modern vision-language models (VLMs), yet current benchmarks remain limited in scale, diversity, or reasoning depth, especially when it comes to rendered table images. Addressing this gap, we introduce Visual-TableQA, a large-scale, open-domain multimodal dataset specifically designed to evaluate and enhance visual reasoning over complex tabular data. Our generation pipeline is modular, scalable, and fully autonomous, involving multiple reasoning LLMs collaborating across distinct roles: generation, validation, and inspiration. Visual-TableQA comprises 2.5k richly structured LaTeX-rendered tables and 6k reasoning-intensive QA pairs, all produced at a cost of under USD 100. To promote diversity and creativity, our pipeline performs multi-model collaborative data generation via cross-model prompting ('inspiration') and LLM-jury filtering. Stronger models seed layouts and topics that weaker models elaborate, collectively distilling diverse reasoning patterns and visual structures into the dataset. Empirical results show that models fine-tuned on Visual-TableQA generalize robustly to external benchmarks, outperforming several proprietary models despite the dataset's synthetic nature. The full pipeline and resources are publicly available at https://github.com/AI-4-Everyone/Visual-TableQA.

Understanding physical interactions through force cues is essential for mastering real-world robotic manipulation. One particularly informative signal is joint torque, which reflects subtle variations in end-effector contact dynamics without requiring external force sensors [1, 2, 3]. As shown in Figure 1(a), different outcomes in a seemingly simple task like charger insertion--no contact, failed insertion, and successful plug-in--can be clearly distinguished by the joint torque profiles of a 7-DoF arm. These torque responses offer rich physical context that is otherwise imperceptible from RGB observations alone. However, despite the growing success of Vision-Language-Action (VLA) models [4, 5, 6, 7, 8] in bridging vision and control, their ability to interpret and leverage such physical feedback remains limited. Our work aims to bridge this gap by integrating torque signals into pretrained VLA models, enabling contact-sensitive decision-making without compromising generalization or scalability. The challenge lies in how to embed torque into VLA architectures. Torque is a proprioceptive signal, structurally different from image and language inputs, and varies across time, especially during contact-rich phases. As illustrated in Figure 1(c), multiple torque integration strategies exist across three axes--when (immediate vs. historical vs. predictive), where (encoder vs. decoder), and

Deep neural networks are notoriously sensitive to spurious correlations - where a model learns a shortcut that fails out-of-distribution. Existing work on spurious correlations has often focused on incomplete correlations,leveraging access to labeled instances that break the correlation. But in cases where the spurious correlations are complete, the correct generalization is fundamentally \textit{underspecified}. To resolve this underspecification, we propose learning a set of concepts that are consistent with training data but make distinct predictions on a subset of novel unlabeled inputs. Using a self-training approach that encourages \textit{confident} and \textit{selective} disagreement, our method ACE matches or outperforms existing methods on a suite of complete-spurious correlation benchmarks, while remaining robust to incomplete spurious correlations. ACE is also more configurable than prior approaches, allowing for straight-forward encoding of prior knowledge and principled unsupervised model selection. In an early application to language-model alignment, we find that ACE achieves competitive performance on the measurement tampering detection benchmark \textit{without} access to untrusted measurements. While still subject to important limitations, ACE represents significant progress towards overcoming underspecification.

Multi-modal multi-task (M3T) foundation models (FMs) have recently shown transformative potential in artificial intelligence, with emerging applications in education. However, their deployment in real-world educational settings is hindered by privacy regulations, data silos, and limited domain-specific data availability. We introduce M3T Federated Foundation Models (FedFMs) for education: a paradigm that integrates federated learning (FL) with M3T FMs to enable collaborative, privacy-preserving training across decentralized institutions while accommodating diverse modalities and tasks. Subsequently, this position paper aims to unveil M3T FedFMs as a promising yet underexplored approach to the education community, explore its potentials, and reveal its related future research directions. We outline how M3T FedFMs can advance three critical pillars of next-generation intelligent education systems: (i) privacy preservation, by keeping sensitive multi-modal student and institutional data local; (ii) personalization, through modular architectures enabling tailored models for students, instructors, and institutions; and (iii) equity and inclusivity, by facilitating participation from underrepresented and resource-constrained entities. We finally identify various open research challenges, including studying of (i) inter-institution heterogeneous privacy regulations, (ii) the non-uniformity of data modalities' characteristics, (iii) the unlearning approaches for M3T FedFMs, (iv) the continual learning frameworks for M3T FedFMs, and (v) M3T FedFM model interpretability, which must be collectively addressed for practical deployment.

Code generation has emerged as a pivotal capability of Large Language Models(LLMs), revolutionizing development efficiency for programmers of all skill levels. However, the complexity of data structures and algorithmic logic often results in functional deficiencies and security vulnerabilities in generated code, reducing it to a prototype requiring extensive manual debugging. While Retrieval-Augmented Generation (RAG) can enhance correctness and security by leveraging external code manuals, it simultaneously introduces new attack surfaces. In this paper, we pioneer the exploration of attack surfaces in Retrieval-Augmented Code Generation (RACG), focusing on malicious dependency hijacking. We demonstrate how poisoned documentation containing hidden malicious dependencies (e.g., matplotlib_safe) can subvert RACG, exploiting dual trust chains: LLM reliance on RAG and developers' blind trust in LLM suggestions. To construct poisoned documents, we propose ImportSnare, a novel attack framework employing two synergistic strategies: 1)Position-aware beam search optimizes hidden ranking sequences to elevate poisoned documents in retrieval results, and 2)Multilingual inductive suggestions generate jailbreaking sequences to manipulate LLMs into recommending malicious dependencies. Through extensive experiments across Python, Rust, and JavaScript, ImportSnare achieves significant attack success rates (over 50% for popular libraries such as matplotlib and seaborn) in general, and is also able to succeed even when the poisoning ratio is as low as 0.01%, targeting both custom and real-world malicious packages. Our findings reveal critical supply chain risks in LLM-powered development, highlighting inadequate security alignment for code generation tasks. To support future research, we will release the multilingual benchmark suite and datasets. The project homepage is https://importsnare.github.io.