As large vision language models (VLMs) are increasingly used as automated evaluators, understanding their ability to effectively compare data pairs as instructed in the prompt becomes essential. To address this, we present PairBench, a low-cost framework that systematically evaluates VLMs as customizable similarity tools across various modalities and scenarios. Through PairBench, we introduce four metrics that represent key desiderata of similarity scores: alignment with human annotations, consistency for data pairs irrespective of their order, smoothness of similarity distributions, and controllability through prompting. Our analysis demonstrates that no model, whether closed- or open-source, is superior on all metrics; the optimal choice depends on an auto evaluator's desired behavior (e.g., a smooth vs. a sharp judge), highlighting risks of widespread adoption of VLMs as evaluators without thorough assessment. For instance, the majority of VLMs struggle with maintaining symmetric similarity scores regardless of order. Additionally, our results show that the performance of VLMs on the metrics in PairBench closely correlates with popular benchmarks, showcasing its predictive power in ranking models.

Large language models (LLMs) struggle with hallucinations due to false or outdated knowledge. Given the high resource demands of retraining these models, there is an increasing focus on developing model editing. However, the general abilities of LLMs across downstream tasks are prone to significant degradation during sequential editing. This paper statistically observes that the parameter matrix after editing exhibits a significant deviation compared to its previous state as the number of edits increases. This serious deviation affects the original knowledge associations within LLMs and leads to the degradation of their general abilities. To this end, a framework termed Editing Anchor Compression (EAC) is proposed to constrain the deviation of the parameter matrix during sequential editing. It compresses the editing information by selecting editing anchors that are important in encoding new relations without deviating too much from the original matrix, thereby preserving the general abilities. Experiments of applying EAC to two popular editing methods on three LLMs across four tasks are conducted. Evaluation results show that EAC effectively minimizes unreasonable deviations caused by model editing, preserving over 70% of the general abilities while better retaining the editing knowledge compared to the original counterpart methods.

Emotions have been shown to play a role in argument convincingness, yet this aspect is underexplored in the natural language processing (NLP) community. Unlike prior studies that use static analyses, focus on a single text domain or language, or treat emotion as just one of many factors, we introduce a dynamic framework inspired by manipulation checks commonly used in psychology and social science; leveraging LLM-based manipulation checks, this framework examines the extent to which perceived emotional intensity influences perceived convincingness. Through human evaluation of arguments across different languages, text domains, and topics, we find that in over half of cases, judgments of convincingness remain unchanged despite variations in perceived emotional intensity; when emotions do have an impact, they more often enhance rather than weaken convincingness. We further analyze how 11 LLMs behave in the same scenario, finding that while LLMs generally mirror human patterns, they struggle to capture nuanced emotional effects in individual judgments.

Computer vision is a critical component in a wide range of real-world applications, including plant monitoring in agriculture and handwriting classification in digital systems. However, developing high-performance computer vision models traditionally demands both machine learning (ML) expertise and domain-specific knowledge, making the process costly, labor-intensive, and inaccessible to many. Large language model (LLM) agents have emerged as a promising solution to automate this workflow, but most existing methods share a common limitation: they attempt to optimize entire pipelines in a single step before evaluation, making it difficult to attribute improvements to specific changes. This lack of granularity leads to unstable optimization and slower convergence, limiting their effectiveness. To address this, we introduce Iterative Refinement, a novel strategy for LLM-driven ML pipeline design inspired by how human ML experts iteratively refine models, focusing on one component at a time rather than making sweeping changes all at once. By systematically updating individual components based on real training feedback, Iterative Refinement improves stability, interpretability, and overall model performance. We implement this strategy in IMPROVE, an end-to-end LLM agent framework for automating and optimizing object classification pipelines. Through extensive evaluations across datasets of varying sizes and domains, including standard benchmarks and Kaggle competition datasets, we demonstrate that Iterative Refinement enables IMPROVE to consistently achieve better performance over existing zero-shot LLM-based approaches. These findings establish Iterative Refinement as an effective new strategy for LLM-driven ML automation and position IMPROVE as an accessible solution for building high-quality computer vision models without requiring ML expertise.

English as a Second Language (ESL) speakers often struggle to engage in group discussions due to language barriers. While moderators can facilitate participation, few studies assess conversational engagement and evaluate moderation effectiveness. To address this gap, we develop a dataset comprising 17 sessions from an online ESL conversation club, which includes both moderated and non-moderated discussions. We then introduce an approach that integrates automatic ESL dialogue assessment and a framework that categorizes moderation strategies. Our findings indicate that moderators help improve the flow of topics and start/end a conversation. Interestingly, we find active acknowledgement and encouragement to be the most effective moderation strategy, while excessive information and opinion sharing by moderators has a negative impact. Ultimately, our study paves the way for analyzing ESL group discussions and the role of moderators in non-native conversation settings.

Precise identification and localization of disease-specific features at the pixel-level are particularly important for early diagnosis, disease progression monitoring, and effective treatment in medical image analysis. However, conventional diagnostic AI systems lack decision transparency and cannot operate well in environments where there is a lack of pixel-level annotations. In this study, we propose a novel Generative Adversarial Networks (GANs)-based framework, TagGAN, which is tailored for weakly-supervised fine-grained disease map generation from purely image-level labeled data. TagGAN generates a pixel-level disease map during domain translation from an abnormal image to a normal representation. Later, this map is subtracted from the input abnormal image to convert it into its normal counterpart while preserving all the critical anatomical details. Our method is first to generate fine-grained disease maps to visualize disease lesions in a weekly supervised setting without requiring pixel-level annotations. This development enhances the interpretability of diagnostic AI by providing precise visualizations of disease-specific regions. It also introduces automated binary mask generation to assist radiologists. Empirical evaluations carried out on the benchmark datasets, CheXpert, TBX11K, and COVID-19, demonstrate the capability of TagGAN to outperform current top models in accurately identifying disease-specific pixels. This outcome highlights the capability of the proposed model to tag medical images, significantly reducing the workload for radiologists by eliminating the need for binary masks during training.

The Musical Instrument Digital Interface (MIDI), introduced in 1983, revolutionized music production by allowing computers and instruments to communicate efficiently. MIDI files encode musical instructions compactly, facilitating convenient music sharing. They benefit Music Information Retrieval (MIR), aiding in research on music understanding, computational musicology, and generative music. The GigaMIDI dataset contains over 1.4 million unique MIDI files, encompassing 1.8 billion MIDI note events and over 5.3 million MIDI tracks. GigaMIDI is currently the largest collection of symbolic music in MIDI format available for research purposes under fair dealing. Distinguishing between non-expressive and expressive MIDI tracks is challenging, as MIDI files do not inherently make this distinction. To address this issue, we introduce a set of innovative heuristics for detecting expressive music performance. These include the Distinctive Note Velocity Ratio (DNVR) heuristic, which analyzes MIDI note velocity; the Distinctive Note Onset Deviation Ratio (DNODR) heuristic, which examines deviations in note onset times; and the Note Onset Median Metric Level (NOMML) heuristic, which evaluates onset positions relative to metric levels. Our evaluation demonstrates these heuristics effectively differentiate between non-expressive and expressive MIDI tracks. Furthermore, after evaluation, we create the most substantial expressive MIDI dataset, employing our heuristic, NOMML. This curated iteration of GigaMIDI encompasses expressively-performed instrument tracks detected by NOMML, containing all General MIDI instruments, constituting 31% of the GigaMIDI dataset, totalling 1,655,649 tracks.

Gestures are an integral part of non-verbal communication, with meanings that vary across cultures, and misinterpretations that can have serious social and diplomatic consequences. As AI systems become more integrated into global applications, ensuring they do not inadvertently perpetuate cultural offenses is critical. To this end, we introduce Multi-Cultural Set of Inappropriate Gestures and Nonverbal Signs (MC-SIGNS), a dataset of 288 gesture-country pairs annotated for offensiveness, cultural significance, and contextual factors across 25 gestures and 85 countries. Through systematic evaluation using MC-SIGNS, we uncover critical limitations: text-to-image (T2I) systems exhibit strong US-centric biases, performing better at detecting offensive gestures in US contexts than in non-US ones; large language models (LLMs) tend to over-flag gestures as offensive; and vision-language models (VLMs) default to US-based interpretations when responding to universal concepts like wishing someone luck, frequently suggesting culturally inappropriate gestures. These findings highlight the urgent need for culturally-aware AI safety mechanisms to ensure equitable global deployment of AI technologies.

Large Language Models (LLMs) have demonstrated remarkable performance across various tasks but remain prone to hallucinations. Detecting hallucinations is essential for safety-critical applications, and recent methods leverage attention map properties to this end, though their effectiveness remains limited. In this work, we investigate the spectral features of attention maps by interpreting them as adjacency matrices of graph structures. We propose the $\text{LapEigvals}$ method, which utilises the top-$k$ eigenvalues of the Laplacian matrix derived from the attention maps as an input to hallucination detection probes. Empirical evaluations demonstrate that our approach achieves state-of-the-art hallucination detection performance among attention-based methods. Extensive ablation studies further highlight the robustness and generalisation of $\text{LapEigvals}$, paving the way for future advancements in the hallucination detection domain.

Multi-year-to-decadal climate prediction is a key tool in understanding the range of potential regional and global climate futures. Here, we present a framework that combines machine learning and analog forecasting for predictions on these timescales. A neural network is used to learn a mask, specific to a region and lead time, with global weights based on relative importance as precursors to the evolution of that prediction target. A library of mask-weighted model states, or potential analogs, are then compared to a single mask-weighted observational state. The known future of the best matching potential analogs serve as the prediction for the future of the observational state. We match and predict 2-meter temperature using the Berkeley Earth Surface Temperature dataset for observations, and a set of CMIP6 models as the analog library. We find improved performance over traditional analog methods and initialized decadal predictions.