Recent advances in self-refinement have demonstrated significant potential for improving the outputs of large language models (LLMs) through iterative refinement. However, most existing self-refinement methods rely on a reactive process with a fixed number of iterations, making it difficult to determine the optimal timing and content of refinement based on the evolving generation context. Inspired by the way humans dynamically refine their thoughts during execution, we propose ProActive Self-Refinement (PASR), a novel method that enables LLMs to refine their outputs during the generation process. Unlike methods that regenerate entire responses, PASR proactively decides whether, when, and how to refine based on the model's internal state and evolving context. We conduct extensive experiments on a diverse set of 10 tasks to evaluate the effectiveness of PASR. Experimental results show that PASR significantly enhances problem-solving performance. In particular, on Qwen3-8B, PASR reduces average token consumption by 41.6% compared to standard generation, while also achieving an 8.2% improvement in accuracy. Our code and baselines used in the paper are available on GitHub.

Abstract--Autonomous driving platforms encounter diverse driving scenarios, each with varying hardware resources and precision requirements. Given the computational limitations of embedded devices, it is crucial to consider computing costs when deploying on target platforms like the DRIVE PX 2. Our objective is to customize the semantic segmentation network according to the computing power and specific scenarios of autonomous driving hardware. We implement dynamic adaptability through a three-tier control mechanism--width multiplier, classifier depth, and classifier kernel--allowing fine-grained control over model components based on hardware constraints and task requirements. This adaptability facilitates broad model scaling, targeted refinement of the final layers, and scenario-specific optimization of kernel sizes, leading to improved resource allocation and performance. Additionally, we leverage Bayesian Optimization with surrogate modeling to efficiently explore hyperparameter spaces under tight computational budgets. It scales its Multiply-Accumulate Operations (MACs) for T ask-Specific Learning Adaptation (TSLA), resulting in alternative configurations tailored to diverse self-driving tasks. These TSLA customizations maximize computational capacity and model accuracy, optimizing hardware utilization. Real-time scene understanding is essential for perception in mobile robotics and self-driving cars. Semantic segmentation, which classifies each pixel in an image into categories like'road' or'sky,' is crucial for assisting in localization and planning for informed decision-making. Semantic segmentation with convolutional neural networks (CNNs) can be computationally expensive, especially with large datasets or complex architectures. Training and deploying CNNs often require significant computational resources, making it a major challenge. Previous research has explored various strategies to address these computational demands. Wang are with the Department of Electrical and Computer Engineering, Northeastern University, Boston, MA, 02115.

Biological synaptic pruning removes weak neural connections to improve efficiency, while standard dropout in artificial networks randomly deactivates neurons without considering connection importance. We propose a magnitude-based synaptic pruning method that better emulates biological processes by gradually removing connections according to their contribution to model performance. Integrated directly into the training loop as a dropout replacement, our method computes weight importance from absolute magnitudes across layers and applies a cubic schedule to progressively increase global sparsity. At regular intervals, pruning masks are updated by thresholding weights, permanently removing low-importance connections while preserving gradient flow for active ones. This continuous, data-driven pruning removes the need for separate pruning and fine-tuning phases. We evaluated the method across multiple time series forecasting architectures, including Recurrent Neural Networks, Long Short-Term Memory, and Patch Time Series Transformer models, using four datasets. Our synaptic pruning approach achieved the best overall performance ranking across all architectures, with statistically significant improvements confirmed by Friedman tests ( p < 0. 01). In financial forecasting tasks, it reduced Mean Absolute Error by up to 20% compared to models using no dropout or standard dropout, with reductions reaching 52% in select transformer models. The proposed mechanism advances regularization by coupling dynamic weight elimination with progressive sparsification during training.

Anomaly detection in cloud environments remains both critical and challenging. Existing context-level benchmarks typically focus on either metrics or logs and often lack reliable annotation, while most detection methods emphasize point anomalies within a single modality, overlooking contextual signals and limiting real-world applicability. Constructing a benchmark for context anomalies that combines metrics and logs is inherently difficult: reproducing anomalous scenarios on real servers is often infeasible or potentially harmful, while generating synthetic data introduces the additional challenge of maintaining cross-modal consistency. Ensuring the stability and availability of large-scale cloud systems is of great importance (Kazemzadeh & Jacobsen, 2009; Bu et al., 2018; Zhang et al., 2015). Accurate detection methods that can also identify among anomaly scenarios are essential to mitigate potential losses (Zhang et al., 2018; Barbhuiya et al., 2018a). Large-scale cloud systems usually generate abundant logs and expose various metrics, both of which serve as some of the most valuable data sources for anomaly detection (Lin et al., 2016; Nandi et al., 2016). Numerous benchmarks have been proposed for cloud anomaly detection such as (Oliner & Stearley, 2007; Xu et al., 2009; Akmeemana et al., 2025). However, most existing research and benchmarks for cloud anomaly detection have focused on point anomalies, where deviations are identified in isolation within a single modality, such as metrics or logs. Although these benchmarks have provided the community with relevant evaluation testbeds, they capture only a narrow slice of the anomaly landscape and often fail to reflect the complexity of real cloud environments.

Once language models (LMs) are deployed, they can interact with users long-term, ideally evolving based on their feedback. Asking for direct user feedback can be disruptive; thus, we study harvesting implicit user feedback from user-LM interaction logs. We study two user-LM interaction datasets (WildChat and LMSYS). First, we analyze user feedback in the user-LLM conversation logs, providing insights into when and why such feedback occurs. Second, we study harvesting learning signals from such implicit user feedback. Specifically, we study whether incorporating the contents of user feedback (e.g., user wanted clarification), in addition to the polarity of the feedback, can improve the model performance. We observe mixed results, showing this helps in short human-designed questions (MTBench) but not on longer and more complex questions (WildBench). Together, we provide an in-depth study of implicit user feedback, showing its potential and limitations.

Abstract--Deep learning achieved great progress recently, however, it is not easy or efficient to further improve its performance by increasing the size of the model. Multi-modal learning can mitigate this challenge by introducing richer and more discriminative information as input. T o solve the problem of limited access to multi-modal data at the time of use, we conduct multi-modal learning by introducing a teacher model to transfer discriminative knowledge to a student model during training. However, this knowledge transfer via distillation is not trivial because the big domain gap between the widely differing modalities can easily lead to overfitting. In this work, we introduce a cross-modal distillation framework. Specifically, we find hard constrained loss, e.g. T o address this, we propose two soft constrained knowledge distillation strategies at the feature level and classifier level respectively . In addition, we propose a quality-based adaptive weights module to weigh input samples via quantified data quality, leading to robust model training. We conducted experiments on speaker recognition and image classification tasks, and the results show that our approach is able to effectively achieve knowledge transfer between the commonly used and widely differing modalities of image, text, and speech. The rapid advancement of deep learning has revolutionized numerous fields by enabling the development of increasingly complex and powerful models. However, as model sizes continue to grow, the marginal benefits of scaling up models diminish, prompting researchers to explore alternative strategies for improving performance. One such strategy is multi-modal learning, which leverages the complementary strengths of multiple data modalities--such as images, speech, and text--to enhance task performance. While multi-modal learning has shown promise in various applications, its practical adoption is often hindered by the high cost and complexity of acquiring and processing multi-modal data. This limitation raises a critical question: how can we effectively utilize multi-modal data during training when only uni-modal data is available during deployment? T o address this challenge, we propose a novel framework for cross-modal knowledge distillation, which enables the transfer of knowledge from a strong modality (e.g., images) to a weak modality (e.g., speech) during training, even when only the weak modality is available during inference.

Recent advancements in multimodal large language models (MLLMs) have driven researchers to explore how well these models read data visualizations, e.g., bar charts, scatter plots. More recently, attention has shifted to visual question answering with maps (Map-VQA). However, Map-VQA research has primarily focused on choropleth maps, which cover only a limited range of thematic categories and visual analytical tasks. To address these gaps, we introduce MapIQ, a benchmark dataset comprising 14,706 question-answer pairs across three map types: choropleth maps, cartograms, and proportional symbol maps spanning topics from six distinct themes (e.g., housing, crime). We evaluate multiple MLLMs using six visual analytical tasks, comparing their performance against one another and a human baseline. An additional experiment examining the impact of map design changes (e.g., altered color schemes, modified legend designs, and removal of map elements) provides insights into the robustness and sensitivity of MLLMs, their reliance on internal geographic knowledge, and potential avenues for improving Map-VQA performance.

The electrocardiogram (ECG) is an essential and effective tool for diagnosing heart diseases. However, its effectiveness can be compromised by noise or unavailability of one or more leads of the standard 12-lead recordings, resulting in diagnostic errors or uncertainty. To address these challenges, we propose TolerantECG, a foundation model for ECG signals that is robust to noise and capable of functioning with arbitrary subsets of the standard 12-lead ECG. TolerantECG training combines contrastive and self-supervised learning frameworks to jointly learn ECG signal representations alongside their corresponding knowledge-retrieval-based text report descriptions and corrupted or lead-missing signals. Comprehensive benchmarking results demonstrate that TolerantECG consistently ranks as the best or second-best performer across various ECG signal conditions and class levels in the PTB-XL dataset, and achieves the highest performance on the MIT-BIH Arrhythmia Database.

The field of AI research is advancing at an unprecedented pace, enabling automated hypothesis generation and experimental design across diverse domains such as biology, mathematics, and artificial intelligence. Despite these advancements, there remains a significant gap in the availability of scalable advising systems capable of providing high-quality, well-reasoned feedback to refine proposed hypotheses and experimental designs. To address this challenge, we explore key factors that underlie the development of robust advising systems, including model size, context length, confidence estimation, and structured reasoning processes. Our findings reveal that a relatively small model, when equipped with a well-compressed literature database and a structured reasoning framework, can outperform powerful general-purpose language models such as Deepseek-R1 in terms of acceptance rates for self-ranked top-30% submissions to ICLR 2025. Moreover, when limited to high-confidence predictions, our system achieves an acceptance rate exceeding 90% on the ICLR 2025 test set, underscoring its potential to significantly enhance the quality and efficiency of hypothesis generation and experimental design. The code is released at https://github.com/HowardLiu0830/GUIDE-Research-Idea-Evaluation.

Diffusion models have achieved remarkable success in generating realistic and versatile images from text prompts. Inspired by the recent advancements of language models, there is an increasing interest in further improving the models by aligning with human preferences. However, we investigate alignment from a divergence minimization perspective and reveal that existing preference optimization methods are typically trapped in suboptimal mean-seeking optimization. In this paper, we introduce Divergence Minimization Preference Optimization (DMPO), a novel and principled method for aligning diffusion models by minimizing reverse KL divergence, which asymptotically enjoys the same optimization direction as original RL. We provide rigorous analysis to justify the effectiveness of DMPO and conduct comprehensive experiments to validate its empirical strength across both human evaluations and automatic metrics. Our extensive results show that diffusion models fine-tuned with DMPO can consistently outperform or match existing techniques, specifically consistently outperforming all baseline models across different base models and test sets, achieving the best PickScore in every case, demonstrating the method's superiority in aligning generative behavior with desired outputs. Overall, DMPO unlocks a robust and elegant pathway for preference alignment, bridging principled theory with practical performance in diffusion models.