As Video Large Multimodal Models (VLMMs) rapidly advance, their inherent complexity introduces significant safety challenges, particularly the issue of mismatched generalization where static safety alignments fail to transfer to dynamic video contexts. We introduce SafeVid, a framework designed to instill video-specific safety principles in VLMMs. SafeVid uniquely transfers robust textual safety alignment capabilities to the video domain by employing detailed textual video descriptions as an interpretive bridge, facilitating LLM-based rule-driven safety reasoning. This is achieved through a closed-loop system comprising: 1) generation of SafeVid-350K, a novel 350,000-pair video-specific safety preference dataset; 2) targeted alignment of VLMMs using Direct Preference Optimization (DPO); and 3) comprehensive evaluation via our new SafeVidBench benchmark. Alignment with SafeVid-350K significantly enhances VLMM safety, with models like LLaVA-NeXT-Video demonstrating substantial improvements (e.g., up to 42.39%) on SafeVidBench. SafeVid provides critical resources and a structured approach, demonstrating that leveraging textual descriptions as a conduit for safety reasoning markedly improves the safety alignment of VLMMs. We have made SafeVid-350K dataset (https://huggingface.co/datasets/yxwang/SafeVid-350K) publicly available.

Large language models (LLMs) exhibit remarkable capabilities in handling natural language tasks; however, they may struggle to consistently follow complex instructions including those involve multiple constraints. Post-training LLMs using supervised fine-tuning (SFT) is a standard approach to improve their ability to follow instructions. In addressing complex instruction following, existing efforts primarily focus on data-driven methods that synthesize complex instruction-output pairs for SFT. However, insufficient attention allocated to crucial sub-contexts may reduce the effectiveness of SFT. In this work, we propose transforming sequentially structured input instruction into multiple parallel instructions containing subcontexts. To support processing this multi-input, we propose MISO (Multi-Input Single-Output), an extension to currently dominant decoder-only transformer-based LLMs. MISO introduces a mixture-of-contexts paradigm that jointly considers the overall instruction-output alignment and the influence of individual sub-contexts to enhance SFT effectiveness. We apply MISO fine-tuning to complex instructionfollowing datasets and evaluate it with standard LLM inference. Empirical results demonstrate the superiority of MISO as a fine-tuning method for LLMs, both in terms of effectiveness in complex instruction-following scenarios and its potential for training efficiency.

General-purpose robots capable of performing diverse tasks require synergistic reasoning and acting capabilities. However, recent dual-system approaches, which separate high-level reasoning from low-level acting, often suffer from challenges such as limited mutual understanding of capabilities between systems and latency issues. This paper introduces OneTwoVLA, a single unified vision-language-action model that can perform both acting (System One) and reasoning (System Two). Crucially, OneTwoVLA adaptively switches between two modes: explicitly reasoning at critical moments during task execution, and generating actions based on the most recent reasoning at other times. To further unlock OneTwoVLA's reasoning and generalization capabilities, we design a scalable pipeline for synthesizing embodied reasoning-centric vision-language data, used for co-training with robot data. We validate OneTwoVLA's effectiveness through extensive experiments, highlighting its superior performance across four key capabilities: long-horizon task planning, error detection and recovery, natural human-robot interaction, and generalizable visual grounding, enabling the model to perform long-horizon, highly dexterous manipulation tasks such as making hotpot or mixing cocktails.

Large Language Models (LLMs) have achieved impressive progress in natural language processing, but their limited ability to retain long-term context constrains performance on document-level or multi-turn tasks. Retrieval-Augmented Generation (RAG) mitigates this by retrieving relevant information from an external corpus. However, existing RAG systems often rely on embedding-based retrieval trained on corpus-level semantic similarity, which can lead to retrieving content that is semantically similar in form but misaligned with the question's true intent. Furthermore, recent RAG variants construct graph- or hierarchy-based structures to improve retrieval accuracy, resulting in significant computation and storage overhead. In this paper, we propose an embedding-free retrieval framework. Our method leverages the logical inferencing ability of LLMs in retrieval using iterative search space refinement guided by our novel importance measure and extend our retrieval results with logically related information without explicit graph construction. Experiments on long-context QA benchmarks, including NovelQA and Marathon, show that our approach outperforms strong baselines while reducing storage and runtime by over an order of magnitude.

Remarkable progress in zero-shot learning (ZSL) has been achieved using generative models. However, existing generative ZSL methods merely generate (imagine) the visual features from scratch guided by the strong class semantic vectors annotated by experts, resulting in suboptimal generative performance and limited scene generalization. To address these and advance ZSL, we propose an inductive variational autoencoder for generative zero-shot learning, dubbed GenZSL. Mimicking human-level concept learning, GenZSL operates by inducting new class samples from similar seen classes using weak class semantic vectors derived from target class names (i.e., CLIP text embedding). To ensure the generation of informative samples for training an effective ZSL classifier, our GenZSL incorporates two key strategies. Firstly, it employs class diversity promotion to enhance the diversity of class semantic vectors. Secondly, it utilizes target class-guided information boosting criteria to optimize the model. Extensive experiments conducted on three popular benchmark datasets showcase the superiority and potential of our GenZSL with significant efficacy and efficiency over f-VAEGAN, e.g., 24.7% performance gains and more than $60\times$ faster training speed on AWA2. Codes are available at https://github.com/shiming-chen/GenZSL.

--The issues of AI risk and AI safety are becoming critical as the prospect of artificial general intelligence (AGI) looms larger . The emergence of extremely large and capable generative models has led to alarming predictions and created a stir from boardrooms to legislatures. As a result, AI alignment has emerged as one of the most important areas in AI research. The goal of this position paper is to argue that the currently dominant vision of AGI in the AI and machine learning (AI/ML) community needs to evolve, and that expectations and metrics for its safety must be informed much more by our understanding of the only existing instance of general intelligence, i.e., the intelligence found in animals, and especially in humans. This change in perspective will lead to a more realistic view of the technology, and allow for better policy decisions. The most successful AI systems today, such as large language models (LLMs) [1]-[5], are based on a computation-alist, statistical, and decision-theoretic paradigm rather than a biological one. As these systems scale up in size, they are improving their performance in areas such as reasoning [6]- [9], and becoming more multimodal [10]-[14]. AI agents [15]- [17], including physical ones [18]-[20], are also becoming increasingly capable. With these rapid advances, there is an expectation that powerful systems with artificial general intelligence (AGI) may soon be at hand. Through all this, there is a general desire that AGI must remain subject to human control and intervention, and must exist only to serve human needs (see, for example, the discussion in [21]). There is also great concern that increasingly powerful AGI systems with autonomous agency might pose serious risks, including existential ones [22]-[27], which has led to a focus on AI alignment, i.e., making AI systems consistent with human norms and preferences [28], [29]. The main position argued in this paper is that: 1) General intelligence should be seen in terms of its archetype: The intelligence of living agents; and 2) The goal of building powerful AGI agents is fundamentally inconsistent with the expectation of complete alignment or near-total control of AGI agents by humans even in principle .

Human preference plays a crucial role in the refinement of large language models (LLMs). However, collecting human preference feedback is costly and most existing datasets neglect the correlation between personalization and preferences. To address this issue, we introduce Fair-PP, a synthetic dataset of personalized preferences targeting social equity, derived from real-world social survey data, which includes 28 social groups, 98 equity topics, and 5 personal preference dimensions. Leveraging GPT-4o-mini, we engage in role-playing based on seven representative persona portrayals guided by existing social survey data, yielding a total of 238,623 preference records. Through Fair-PP, we also contribute (i) An automated framework for generating preference data, along with a more fine-grained dataset of personalized preferences; (ii) analysis of the positioning of the existing mainstream LLMs across five major global regions within the personalized preference space; and (iii) a sample reweighting method for personalized preference alignment, enabling alignment with a target persona while maximizing the divergence from other personas. Empirical experiments show our method outperforms the baselines.

Large reasoning models, often post-trained on long chain-of-thought (long CoT) data with reinforcement learning, achieve state-of-the-art performance on mathematical, coding, and domain-specific reasoning benchmarks. However, their logical reasoning capabilities - fundamental to human cognition and independent of domain knowledge - remain understudied. To address this gap, we introduce LogiEval, a holistic benchmark for evaluating logical reasoning in large reasoning models. LogiEval spans diverse reasoning types (deductive, inductive, analogical, and abductive) and task formats (e.g., logical sequence, argument analysis), sourced from high-quality human examinations (e.g., LSAT, GMAT). Our experiments demonstrate that modern reasoning models excel at 4-choice argument analysis problems and analogical reasoning, surpassing human performance, yet exhibit uneven capabilities across reasoning types and formats, highlighting limitations in their generalization. Our analysis reveals that human performance does not mirror model failure distributions. To foster further research, we curate LogiEval-Hard, a challenging subset identified through a novel screening paradigm where small-model failures (Qwen3-30B-A3B) reliably predict difficulties for larger models. Modern models show striking, consistent failures on LogiEval-Hard. This demonstrates that fundamental reasoning bottlenecks persist across model scales, and establishes LogiEval-Hard as both a diagnostic tool and a rigorous testbed for advancing logical reasoning in LLMs.

Robustness against uncertain and ambiguous inputs is a critical challenge for deep learning models. While recent advancements in large scale vision language models (VLMs, e.g. GPT4o) might suggest that increasing model and training dataset size would mitigate this issue, our empirical evaluation shows a more complicated picture. Testing models using two classic uncertainty quantification tasks, anomaly detection and classification under inherently ambiguous conditions, we find that newer and larger VLMs indeed exhibit improved robustness compared to earlier models, but still suffer from a tendency to strictly follow instructions, often causing them to hallucinate confident responses even when faced with unclear or anomalous inputs. Remarkably, for natural images such as ImageNet, this limitation can be overcome without pipeline modifications: simply prompting models to abstain from uncertain predictions enables significant reliability gains, achieving near-perfect robustness in several settings. However, for domain-specific tasks such as galaxy morphology classification, a lack of specialized knowledge prevents reliable uncertainty estimation. Finally, we propose a novel mechanism based on caption diversity to reveal a model's internal uncertainty, enabling practitioners to predict when models will successfully abstain without relying on labeled data.

Large language models (LLMs) have shown remarkable progress in mathematical problem-solving, but evaluation has largely focused on problems that have exact analytical solutions or involve formal proofs, often overlooking approximation-based problems ubiquitous in applied science and engineering. To fill this gap, we build on prior work and present HARDMath2, a dataset of 211 original problems covering the core topics in an introductory graduate applied math class, including boundary-layer analysis, WKB methods, asymptotic solutions of nonlinear partial differential equations, and the asymptotics of oscillatory integrals. This dataset was designed and verified by the students and instructors of a core graduate applied mathematics course at Harvard. We build the dataset through a novel collaborative environment that challenges students to write and refine difficult problems consistent with the class syllabus, peer-validate solutions, test different models, and automatically check LLM-generated solutions against their own answers and numerical ground truths. Evaluation results show that leading frontier models still struggle with many of the problems in the dataset, highlighting a gap in the mathematical reasoning skills of current LLMs. Importantly, students identified strategies to create increasingly difficult problems by interacting with the models and exploiting common failure modes. This back-and-forth with the models not only resulted in a richer and more challenging benchmark but also led to qualitative improvements in the students' understanding of the course material, which is increasingly important as we enter an age where state-of-the-art language models can solve many challenging problems across a wide domain of fields.