Language models are increasingly being incorporated as components in larger AI systems for various purposes, from prompt optimization to automatic evaluation. In this work, we analyze the construct validity of four recent, commonly used methods for measuring text-to-image consistency - CLIPScore, TIFA, VPEval, and DSG - which rely on language models and/or VQA models as components. We define construct validity for text-image consistency metrics as a set of desiderata that text-image consistency metrics should have, and find that no tested metric satisfies all of them. We find that metrics lack sufficient sensitivity to language and visual properties. Next, we find that TIFA, VPEval and DSG contribute novel information above and beyond CLIPScore, but also that they correlate highly with each other. We also ablate different aspects of the text-image consistency metrics and find that not all model components are strictly necessary, also a symptom of insufficient sensitivity to visual information. Finally, we show that all three VQA-based metrics likely rely on familiar text shortcuts (such as yes-bias in QA) that call their aptitude as quantitative evaluations of model performance into question.

Recent advances in language modeling demonstrate the need for high-quality domain-specific training data, especially for tasks that require specialized knowledge. General-purpose models, while versatile, often lack the depth needed for expert-level tasks because of limited domain-specific information. Domain adaptation training can enhance these models, but it demands substantial, high-quality data. To address this, we propose ORBIT, a cost-efficient methodology for curating massive, high-quality domain-specific datasets from noisy web sources, tailored for training specialist large language models. Using astronomy as a primary case study, we refined the 1.3T-token FineWeb-Edu dataset into a high-quality, 10B-token subset focused on astronomy. Fine-tuning \textsc{LLaMA-3-8B} on a 1B-token astronomy subset improved performance on the MMLU astronomy benchmark from 69\% to 76\% and achieved top results on AstroBench, an astronomy-specific benchmark. Moreover, our model (Orbit-LLaMA) outperformed \textsc{LLaMA-3-8B-base}, with GPT-4o evaluations preferring it in 73\% of cases across 1000 astronomy-specific questions. Additionally, we validated ORBIT's generalizability by applying it to law and medicine, achieving a significant improvement of data quality compared to an unfiltered baseline. We open-source the ORBIT methodology, including the curated datasets, the codebase, and the resulting model at \href{https://github.com/ModeEric/ORBIT-Llama}{https://github.com/ModeEric/ORBIT-Llama}.

To facilitate human--robot interaction (HRI) tasks in real-world scenarios, service robots must adapt to dynamic environments and understand the required tasks while effectively communicating with humans. To accomplish HRI in practice, we propose a novel indoor dynamic map, task understanding system, and response generation system. The indoor dynamic map optimizes robot behavior by managing an occupancy grid map and dynamic information, such as furniture and humans, in separate layers. The task understanding system targets tasks that require multiple actions, such as serving ordered items. Task representations that predefine the flow of necessary actions are applied to achieve highly accurate understanding. The response generation system is executed in parallel with task understanding to facilitate smooth HRI by informing humans of the subsequent actions of the robot. In this study, we focused on waiter duties in a restaurant setting as a representative application of HRI in a dynamic environment. We developed an HRI system that could perform tasks such as serving food and cleaning up while communicating with customers. In experiments conducted in a simulated restaurant environment, the proposed HRI system successfully communicated with customers and served ordered food with 90\% accuracy. In a questionnaire administered after the experiment, the HRI system of the robot received 4.2 points out of 5. These outcomes indicated the effectiveness of the proposed method and HRI system in executing waiter tasks in real-world environments.

Recent AI-driven step-function advances in several longstanding problems in music technology are opening up new avenues to create the next generation of music education tools. Creating personalized, engaging, and effective learning experiences is a continuously evolving challenge in music education. Here we present two case studies using such advances in music technology to address these challenges. In our first case study we showcase an application that uses Automatic Chord Recognition to generate personalized exercises from audio tracks, connecting traditional ear training with real-world musical contexts. In the second case study we prototype adaptive piano method books that use Automatic Music Transcription to generate exercises at different skill levels while retaining a close connection to musical interests. These applications demonstrate how recent AI developments can democratize access to high-quality music education and promote rich interaction with music in the age of generative AI. We hope this work inspires other efforts in the community, aimed at removing barriers to access to high-quality music education and fostering human participation in musical expression.

Since the debut of DPO, it has been shown that aligning a target LLM with human preferences via the KL-constrained RLHF loss is mathematically equivalent to a special kind of reward modeling task. Concretely, the task requires: 1) using the target LLM to parameterize the reward model, and 2) tuning the reward model so that it has a 1:1 linear relationship with the true reward. However, we identify a significant issue: the DPO loss might have multiple minimizers, of which only one satisfies the required linearity condition. The problem arises from a well-known issue of the underlying Bradley-Terry preference model: it does not always have a unique maximum likelihood estimator (MLE). Consequently,the minimizer of the RLHF loss might be unattainable because it is merely one among many minimizers of the DPO loss. As a better alternative, we propose an energy-based model (EBM) that always has a unique MLE, inherently satisfying the linearity requirement. To approximate the MLE in practice, we propose a contrastive loss named Energy Preference Alignment (EPA), wherein each positive sample is contrasted against one or more strong negatives as well as many free weak negatives. Theoretical properties of our EBM enable the approximation error of EPA to almost surely vanish when a sufficient number of negatives are used. Empirically, we demonstrate that EPA consistently delivers better performance on open benchmarks compared to DPO, thereby showing the superiority of our EBM.

Knowledge distillation (KD) has become a prevalent technique for compressing large language models (LLMs). Existing KD methods are constrained by the need for identical tokenizers (i.e., vocabularies) between teacher and student models, limiting their versatility in handling LLMs of different architecture families. In this paper, we introduce the Multi-Level Optimal Transport (MultiLevelOT), a novel approach that advances the optimal transport for universal cross-tokenizer knowledge distillation. Our method aligns the logit distributions of the teacher and the student at both token and sequence levels using diverse cost matrices, eliminating the need for dimensional or token-by-token correspondence. At the token level, MultiLevelOT integrates both global and local information by jointly optimizing all tokens within a sequence to enhance robustness. At the sequence level, we efficiently capture complex distribution structures of logits via the Sinkhorn distance, which approximates the Wasserstein distance for divergence measures. Extensive experiments on tasks such as extractive QA, generative QA, and summarization demonstrate that the MultiLevelOT outperforms state-of-the-art cross-tokenizer KD methods under various settings. Our approach is robust to different student and teacher models across model families, architectures, and parameter sizes.

Medical video generation has transformative potential for enhancing surgical understanding and pathology insights through precise and controllable visual representations. However, current models face limitations in controllability and authenticity. To bridge this gap, we propose SurgSora, a motion-controllable surgical video generation framework that uses a single input frame and user-controllable motion cues. SurgSora consists of three key modules: the Dual Semantic Injector (DSI), which extracts object-relevant RGB and depth features from the input frame and integrates them with segmentation cues to capture detailed spatial features of complex anatomical structures; the Decoupled Flow Mapper (DFM), which fuses optical flow with semantic-RGB-D features at multiple scales to enhance temporal understanding and object spatial dynamics; and the Trajectory Controller (TC), which allows users to specify motion directions and estimates sparse optical flow, guiding the video generation process. The fused features are used as conditions for a frozen Stable Diffusion model to produce realistic, temporally coherent surgical videos. Extensive evaluations demonstrate that SurgSora outperforms state-of-the-art methods in controllability and authenticity, showing its potential to advance surgical video generation for medical education, training, and research. See our project page for more results: surgsora.github.io.

As artificial intelligence (AI) continues to advance, understanding public perceptions -- including biases, risks, and benefits -- is critical for guiding research priorities, shaping public discourse, and informing policy. This study explores public mental models of AI using micro scenarios to assess reactions to 71 statements about AI's potential future impacts. Drawing on cross-cultural samples from Germany (N=52) and China (N=60), we identify significant differences in expectations, evaluations, and risk-utility tradeoffs. German participants tended toward more cautious assessments, whereas Chinese participants expressed greater optimism regarding AI's societal benefits. Chinese participants exhibited relatively balanced risk-benefit tradeoffs ($\beta=-0.463$ for risk and $\beta=+0.484$ for benefit, $r^2=.630$). In contrast, German participants showed a stronger emphasis on AI benefits and less on risks ($\beta=-0.337$ for risk and $\beta=+0.715$ for benefit, $r^2=.839$). Visual cognitive maps illustrate these contrasts, offering new perspectives on how cultural contexts shape AI acceptance. Our findings underline key factors influencing public perception and provide actionable insights for fostering equitable and culturally sensitive integration of AI technologies.

Large language models (LLMs) have revolutionized video-based computer vision applications, including action recognition, anomaly detection, and video summarization. Videos inherently pose unique challenges, combining spatial complexity with temporal dynamics that are absent in static images or textual data. Current approaches to video understanding with LLMs often rely on pretrained video encoders to extract spatiotemporal features and text encoders to capture semantic meaning. These representations are integrated within LLM frameworks, enabling multimodal reasoning across diverse video tasks. However, the critical question persists: Can LLMs truly understand the concept of time, and how effectively can they reason about temporal relationships in videos? This work critically examines the role of LLMs in video processing, with a specific focus on their temporal reasoning capabilities. We identify key limitations in the interaction between LLMs and pretrained encoders, revealing gaps in their ability to model long-term dependencies and abstract temporal concepts such as causality and event progression. Furthermore, we analyze challenges posed by existing video datasets, including biases, lack of temporal annotations, and domain-specific limitations that constrain the temporal understanding of LLMs. To address these gaps, we explore promising future directions, including the co-evolution of LLMs and encoders, the development of enriched datasets with explicit temporal labels, and innovative architectures for integrating spatial, temporal, and semantic reasoning. By addressing these challenges, we aim to advance the temporal comprehension of LLMs, unlocking their full potential in video analysis and beyond.

Tree-based learning methods such as Random Forest and XGBoost are still the gold-standard prediction methods for tabular data. Feature importance measures are usually considered for feature selection as well as to assess the effect of features on the outcome variables in the model. This also applies to survey data, which are frequently encountered in the social sciences and official statistics. These types of datasets often present the challenge of missing values. The typical solution is to impute the missing data before applying the learning method. However, given the large number of possible imputation methods available, the question arises as to which should be chosen to achieve the 'best' reflection of feature importance and feature selection in subsequent analyses. In the present paper, we investigate this question in a survey-based simulation study for eight state-of-the art imputation methods and three learners. The imputation methods comprise listwise deletion, three MICE options, four \texttt{missRanger} options as well as the recently proposed mixGBoost imputation approach. As learners, we consider the two most common tree-based methods, Random Forest and XGBoost, and an interpretable linear model with regularization.