Recent advancements in reinforcement learning (RL) have enhanced the reasoning abilities of large language models (LLMs), yet the impact on multimodal LLMs (MLLMs) is limited. Particularly in vision-intensive tasks like geometric reasoning, MLLMs hallucinate frequently, leading to inaccurate reasoning. We attribute this to the perceptual bottleneck in MLLMs, which caps the benefits of reasoning training. To quantify this, we design a Geo-Perception Question-Answering (GeoPQA) benchmark, targeting basic geometric concepts and spatial relationships. Experiments on GeoPQA reveal significant shortcomings of MLLMs in visual perception, which constrain RL reward signals for effective training. To address this bottleneck, we propose a two-stage RL training framework by first enhancing the visual perception of geometric structures, then fostering reasoning capabilities. Applied to Qwen2.5-VL-3B-Instruct, our two-stage training improves geometric reasoning by 9.7% and geometric problem solving by 9.1%, compared to the direct reasoning training approach. Our method also generalizes to other vision-intensive domains like figure understanding, highlighting the importance of perceptual grounding in effective MLLM reasoning.

While various multimodal multi-image evaluation datasets have been emerged, but these datasets are primarily based on English, and there has yet to be a Chinese multi-image dataset. To fill this gap, we introduce RealBench, the first Chinese multimodal multi-image dataset, which contains 9393 samples and 69910 images. RealBench distinguishes itself by incorporating real user-generated content, ensuring high relevance to real-world applications. Additionally, the dataset covers a wide variety of scenes, image resolutions, and image structures, further increasing the difficulty of multi-image understanding. Ultimately, we conduct a comprehensive evaluation of RealBench using 21 multimodal LLMs of different sizes, including closed-source models that support multi-image inputs as well as open-source visual and video models. The experimental results indicate that even the most powerful closed-source models still face challenges when handling multi-image Chinese scenarios. Moreover, there remains a noticeable performance gap of around 71.8\% on average between open-source visual/video models and closed-source models. These results show that RealBench provides an important research foundation for further exploring multi-image understanding capabilities in the Chinese context.

Spelling correction from visual input poses unique challenges for vision language models (VLMs), as it requires not only detecting but also correcting textual errors directly within images. We present ReViCo (Real Visual Correction), the first benchmark that systematically evaluates VLMs on real-world visual spelling correction across Chinese and English. ReViCo contains naturally occurring errors collected from real-world image data and supports fine-grained evaluation at both image and token levels. Through comprehensive experiments on representative cascaded (Qwen) and native (InternVL) open-source models, as well as closed-source systems (GPT-4o, Claude), we show that current VLMs fall significantly short of human performance, particularly in correction. To address these limitations, we explore two solution paradigms: a Joint OCR-Correction pipeline and a Background Information enhanced approach, both of which yield consistent performance gains. Our analysis highlights fundamental limitations of existing architectures and provides actionable insights for advancing multimodal spelling correction.

Ensuring the safety of neural networks under input uncertainty is a fundamental challenge in safety-critical applications. This paper builds on and expands Fazlyab's quadratic-constraint (QC) and semidefinite-programming (SDP) framework for neural network verification to a distributionally robust and tail-risk-aware setting by integrating worst-case Conditional Value-at-Risk (WC-CVaR) over a moment-based ambiguity set with fixed mean and covariance. The resulting conditions remain SDP-checkable and explicitly account for tail risk. This integration broadens input-uncertainty geometry-covering ellipsoids, polytopes, and hyperplanes-and extends applicability to safety-critical domains where tail-event severity matters. Applications to closed-loop reachability of control systems and classification are demonstrated through numerical experiments, illustrating how the risk level $\varepsilon$ trades conservatism for tolerance to tail events-while preserving the computational structure of prior QC/SDP methods for neural network verification and robustness analysis.

Indonesia's marine ecosystems, part of the globally recognized Coral Triangle, are among the richest in biodiversity, requiring efficient monitoring tools to support conservation. Traditional fish detection methods are time-consuming and demand expert knowledge, prompting the need for automated solutions. This study explores the implementation of YOLOv10-nano, a state-of-the-art deep learning model, for real-time marine fish detection in Indonesian waters, using test data from Bunaken National Marine Park. YOLOv10's architecture, featuring improvements like the CSPNet backbone, PAN for feature fusion, and Pyramid Spatial Attention Block, enables efficient and accurate object detection even in complex environments. The model was evaluated on the DeepFish and OpenImages V7-Fish datasets. Results show that YOLOv10-nano achieves a high detection accuracy with mAP50 of 0.966 and mAP50:95 of 0.606 while maintaining low computational demand (2.7M parameters, 8.4 GFLOPs). It also delivered an average inference speed of 29.29 FPS on the CPU, making it suitable for real-time deployment. Although OpenImages V7-Fish alone provided lower accuracy, it complemented DeepFish in enhancing model robustness. Overall, this study demonstrates YOLOv10-nano's potential for efficient, scalable marine fish monitoring and conservation applications in data-limited environments.

How do vision transformers (ViTs) represent and process the world? This paper addresses this long-standing question through the first systematic analysis of 6.6K features across all layers, extracted via sparse autoencoders, and by introducing the residual replacement model, which replaces ViT computations with interpretable features in the residual stream. Our analysis reveals not only a feature evolution from low-level patterns to high-level semantics, but also how ViTs encode curves and spatial positions through specialized feature types. The residual replacement model scalably produces a faithful yet parsimonious circuit for human-scale interpretability by significantly simplifying the original computations. As a result, this framework enables intuitive understanding of ViT mechanisms. Finally, we demonstrate the utility of our framework in debiasing spurious correlations.

Anomaly detection in dynamic graphs is a critical task with broad real-world applications, including social networks, e-commerce, and cybersecurity. Most existing methods assume that normal patterns remain stable over time; however, this assumption often fails in practice due to the phenomenon we refer to as normality distribution shift (NDS), where normal behaviors evolve over time. Ignoring NDS can lead models to misclassify shifted normal instances as anomalies, degrading detection performance. To tackle this issue, we propose WhENDS, a novel unsupervised anomaly detection method that aligns normal edge embeddings across time by estimating distributional statistics and applying whitening transformations. Extensive experiments on four widely-used dynamic graph datasets show that WhENDS consistently outperforms nine strong baselines, achieving state-of-the-art results and underscoring the importance of addressing NDS in dynamic graph anomaly detection.

We introduce FinDebate, a multi-agent framework for financial analysis, integrating collaborative debate with domain-specific Retrieval-Augmented Generation (RAG). Five specialized agents, covering earnings, market, sentiment, valuation, and risk, run in parallel to synthesize evidence into multi-dimensional insights. To mitigate overconfidence and improve reliability, we introduce a safe debate protocol that enables agents to challenge and refine initial conclusions while preserving coherent recommendations. Experimental results, based on both LLM-based and human evaluations, demonstrate the framework's efficacy in producing high-quality analysis with calibrated confidence levels and actionable investment strategies across multiple time horizons.

LLMs suffer from critical reasoning issues such as unfaithfulness, bias, and inconsistency, since they lack robust causal underpinnings and may rely on superficial correlations rather than genuine understanding. Successive LRMs have emerged as a promising alternative, leveraging advanced training techniques such as reinforcement learning (RL) and distillation to improve task accuracy. However, the impact of these training methods on causality remains largely unexplored. In this study, we conduct a systematic causal analysis on LLMs and LRMs, examining structural causal models (SCMs) of four key variables: problem instruction (Z), thinking process (T), reasoning steps (X), and answer (Y). Our findings reveal that RLVR-trained LRMs exhibit enhanced causal reasoning capabilities, aligning more closely with ideal causal structures, while LLMs and distilled LRMs fail to address causality-related deficiencies. Our further investigation indicates that RLVR reduces spurious correlations and strengthens genuine causal patterns, thereby mitigating unfaithfulness and bias. In addition, our inspection on the dynamics of the RLVR training process observes a high correlation between reduced spurious features and improved causal structures, where the causal relationships consistently improve in the training process. This study contributes to the understanding of causality in reasoning models, highlights the critical role of RLVR in enhancing causal reasoning, and provides insights for designing future AI systems with stronger causal foundations. We release our code and data at https://github.com/Harryking1999/CoT_Causal_Analysis.

We present a comparative analysis of text complexity across domains using scale-free metrics. We quantify linguistic complexity via Heaps' exponent $β$ (vocabulary growth), Taylor's exponent $α$ (word-frequency fluctuation scaling), compression rate $r$ (redundancy), and entropy. Our corpora span three domains: legal documents (statutes, cases, deeds) as a specialized domain, general natural language texts (literature, Wikipedia), and AI-generated (GPT) text. We find that legal texts exhibit slower vocabulary growth (lower $β$) and higher term consistency (higher $α$) than general texts. Within legal domain, statutory codes have the lowest $β$ and highest $α$, reflecting strict drafting conventions, while cases and deeds show higher $β$ and lower $α$. In contrast, GPT-generated text shows the statistics more aligning with general language patterns. These results demonstrate that legal texts exhibit domain-specific structures and complexities, which current generative models do not fully replicate.