Preference optimization (PO) is indispensable for large language models (LLMs), with methods such as direct preference optimization (DPO) and proximal policy optimization (PPO) achieving great success. A common belief is that DPO is supervised learning while PPO is reinforcement learning, yet deeper analyses for the reasons underlying these differences remain lacking. To fill this gap, we analyze their optimization dynamics, revealing distinct algorithmic behaviors and comprehending their underlying causes. First, we examine the target directions of gradient-based updates and find that DPO follows stable targets, whereas PPO follows dynamic targets that balance exploration and exploitation, thus validating the common belief from a new perspective. Second, we examine the roles of positive learning, negative learning, and loss reweighting, which are three key components in PO methods. Our analyses reveal that these components play fairly different roles. In DPO, positive and negative learning jointly shape the learning targets meanwhile mutually offset each other. However, loss reweighting in DPO acts less as a reward signal but more as a regularizer to mitigate overfitting. In PPO, negative learning primarily supports exploration rather than determining the targets. Meanwhile, loss reweighting, related to absolute values of token-level advantages, indicates the distinct roles of token groups in updating targets. Given these findings, we conduct carefully designed ablation studies to further examine how controlling these dynamics impacts optimization efficiency and practical performance. The insights gained from our analyses not only deepen the understanding of PO methods but also inspire the development of more preference-aligned LLMs.

Retrieval-Augmented Generation (RAG) is gaining recognition as one of the key technological axes for next generation information retrieval, owing to its ability to mitigate the hallucination phenomenon in Large Language Models (LLMs)and effectively incorporate up-to-date information. However, specialized expertise is necessary to construct ahigh-quality retrieval system independently; moreover, RAGdemonstratesrelativelyslowerprocessing speeds compared to conventional pure retrieval systems because it involves both retrieval and generation stages. Accordingly, this study proposes SHRAG, a novel framework designed to facilitate the seamless integration of Information Retrieval and RAG while simultaneously securing precise retrieval performance. SHRAG utilizes a Large Language Model as a Query Strategist to automatically transform unstructured natural language queries into logically structured search queries, subsequently performing Boolean retrieval to emulate the search process of an expert human searcher. Furthermore, it incorporates multilingual query expansion and a multilingual embedding model, enabling it to perform efficient cross-lingual question answering within the multilingual dataset environment of the ScienceON Challenge. Experimental results demonstrate that the proposed method, combining logical retrieval capabilities and generative reasoning, can significantly enhance the accuracy and reliability of RAG systems. Furthermore, SHRAG movesbeyondconventionaldocument-centric retrieval methods, presenting the potential for a new search paradigm capable of providing direct and reliable responses to queries.

Adaptive optimization methods (such as Adam) play a major role in LLM pretraining, significantly outperforming Gradient Descent (GD). Recent studies have proposed new smoothness assumptions on the loss function to explain the advantages of adaptive algorithms with structured preconditioners, e.g., coordinate-wise or layer-wise, and steepest descent methods w.r.t. non-euclidean norms, e.g., $\ell_\infty$ norm or spectral norm, over GD. However, it remains unclear how these smoothness assumptions manifest in language modelling tasks. In this work, we aim to analyze the benefit of $\ell_\infty$-norm descent (a.k.a. sign descent) directly from properties of the data distribution, namely, heavy-tailed class imbalance. We propose a minimal yet representative setting of next-token prediction, where we can provably show faster convergence of coordinate-wise algorithms such as Sign descent (steepest descent w.r.t. $\ell_\infty$ norm) over normalized GD (steepest descent w.r.t. to $\ell_2$ norm) in the presence of heavy tail class imbalance.

With the advancement of computational resources, Large Vision-Language Models (LVLMs) exhibit impressive Perception and Reasoning (P&R) performance on Graphical User Interface (GUI) tasks. However, although they demonstrate strong P&R capabilities in English GUI scenarios, their performance in multilingual settings has received little attention, which limits their global applications. Moreover, existing studies on GUI tasks lack fine-grained analyses, including widget functions and elements' spatial relationships, which are fundamental for more targeted improvements. To tackle these issues, we propose MPR-GUI-Bench, a Multilingual fine-grained Perception and Reasoning GUI Benchmark to evaluate GUI agents' P&R capabilities. Evaluation results demonstrate that LVLMs exhibit significantly worse P&R performance in non-English languages than in English. To address these gaps, we propose GUI-XLI, a GUI Cross-Lingual Intervention method that applies interventions to the hidden states at P&R capability-related layers to mitigate the gaps between English and other languages, building on previous research showing that the hidden states of different language inputs exhibit significant differences in the latent space. Experimental results indicate that our method improves GUI agents' multilingual P&R capability by 6.5% on average.

Abstract--In recent years, the explosion of malware and extensive code reuse have formed complex evolutionary connections among malware specimens. The rapid pace of development makes it challenging for existing studies to characterize recent evolutionary trends. In addition, intuitive tools to untangle these intricate connections between malware specimens or categories are urgently needed. This paper introduces a manually-reviewed malware source code dataset containing 6032 specimens. Building on and extending current research from a software engineering perspective, we systematically evaluate the scale, development costs, code quality, as well as security and dependencies of modern malware. We further introduce a multi-view genealogy analysis to clarify malware connections: at an overall view, this analysis quantifies the strength and direction of connections among specimens and categories; at a detailed view, it traces the evolutionary histories of individual specimens. Experimental results indicate that, despite persistent shortcomings in code quality, malware specimens exhibit an increasing complexity and standardization, in step with the development of mainstream software engineering practices. Meanwhile, our genealogy analysis intuitively reveals lineage expansion and evolution driven by code reuse, providing new evidence and tools for understanding the formation and evolution of the malware ecosystem. With the rapid development of information technology and large language models, malware has experienced a surge in recent years, exhibiting strong connections among categories and specimens, as well as high code reuse rates [1]. In the past 12 months, more than 107 million new malicious or potentially unwanted applications were detected [2], [3]. Many of these malware specimens are variants of previously known malware, which indicates the prevalence of code reuse and family-oriented evolution. However, the difficulty of collecting, reviewing, and labeling has resulted in a scarcity of source code datasets [4]. Existing datasets lack human curation, reliable labels, and timestamps.

Humans build viewpoint-independent cognitive maps through navigation, enabling intuitive reasoning about object permanence and spatial relations. We argue that multimodal large language models (MLLMs), despite extensive video training, lack this fundamental spatial reasoning capability, a critical limitation for embodied applications. To demonstrate these limitations and drive research, we introduce REM (Reasoning over Embodied Multi-Frame Trajectories), a benchmark using controllable 3D environments for long-horizon embodied spatial reasoning. REM systematically evaluates key aspects like object permanence/distinction, spatial relationships, and numerical tracking across dynamic embodied viewpoints. Our evaluation shows that the best-performing current models exhibit promising overall performance, but become increasingly unreliable at even moderate complexity levels easily handled by humans. These findings highlight challenges MLLMs face in developing robust spatial representations from sequential visual input. Consequently, REM provides targeted metrics and diagnostics to foster improved spatial understanding in future models.

Large reasoning models (LRMs) have garnered significant attention from researchers owing to their exceptional capability in addressing complex tasks. Motivated by the observed human-like behaviors in their reasoning processes, this paper introduces a comprehensive taxonomy to characterize atomic reasoning steps and probe the ``psyche'' of LRM intelligence. Specifically, it comprises five groups and seventeen categories derived from human mental processes, thereby grounding the understanding of LRMs in an interdisciplinary perspective. The taxonomy is then applied for an in-depth understanding of current LRMs, resulting in a distinct labeled dataset that comprises 277,534 atomic reasoning steps. Using this resource, we analyze contemporary LRMs and distill several actionable takeaways for improving training and post-training of reasoning models. Notably, our analysis reveals that prevailing post-answer ``double-checks'' (self-monitoring evaluations) are largely superficial and rarely yield substantive revisions. Thus, incentivizing comprehensive multi-step reflection, rather than simple self-monitoring, may offer a more effective path forward. To complement the taxonomy, an automatic annotation framework, named CAPO, is proposed to leverage large language models (LLMs) for generating the taxonomy-based annotations. Experimental results demonstrate that CAPO achieves higher consistency with human experts compared to baselines, facilitating a scalable and comprehensive analysis of LRMs from a human cognitive perspective. Together, the taxonomy, CAPO, and the derived insights provide a principled, scalable path toward understanding and advancing LRM reasoning.

Typical deep clustering methods, while achieving notable progress, can only provide one clustering result per dataset. This limitation arises from their assumption of a fixed underlying data distribution, which may fail to meet user needs and provide unsatisfactory clustering outcomes. Our work investigates how multi-modal large language models (MLLMs) can be leveraged to achieve user-driven clustering, emphasizing their adaptability to user-specified semantic requirements. However, directly using MLLM output for clustering has risks for producing unstructured and generic image descriptions instead of feature-specific and concrete ones. To address these issues, our method first discovers that MLLMs' hidden states of text tokens are strongly related to the corresponding features, and leverages these embeddings to perform clusterings from any user-defined criteria. We also employ a lightweight clustering head augmented with pseudo-label learning, significantly enhancing clustering accuracy. Extensive experiments demonstrate its competitive performance on diverse datasets and metrics.

In this paper, we point out that the objective of the retrieval algorithms is to align with the LLM, which is similar to the objective of knowledge distillation in LLMs. We analyze the similarity in information focus between the distilled language model(DLM) and the original LLM from the perspective of information theory, and thus propose a novel paradigm that leverages a DLM as the retrieval algorithm. Based on the insight, we present SpeContext, an algorithm and system co-design for long-context reasoning. (1) At the algorithm level, SpeContext proposes lightweight retrieval head based on the head-level attention weights of DLM, achieving > 90% parameters reduction by pruning the redundancy. (2) At the system level, SpeContext designs an asynchronous prefetch dataflow via the elastic loading strategy, effectively overlapping KV cache retrieval with the LLM computation. (3) At the compilation level, SpeContext constructs the theoretical memory model and implements an adaptive memory management system to achieve acceleration by maximizing GPU memory utilization. We deploy and evaluate SpeContext in two resourceconstrained environments, cloud and edge. Extensive experiments show that, compared with the Huggingface framework, SpeContext achieves up to 24.89x throughput improvement in cloud and 10.06x speedup in edge with negligible accuracy loss, pushing the Pareto frontier of accuracy and throughput.

Quality of datasets plays an important role in large language model (LLM) alignment. In collecting human feedback, however, preference flipping is ubiquitous and causes corruption in data annotation; the issue necessitates the alignment algorithms with improved robustness against potential flipped pairs. To this end, this paper introduces a Flipping-Aware Direct Preference Optimization (FA-DPO) algorithm tailored to preference flipping from a reinforcement learning with human feedback (RLHF) perspective. We dissect the inherent human intention model and the preference flipping mechanism introduced by external factors as two distinct stages; in the latter, we introduce an instance-dependent flipping probability on the basis of the Bradley-Terry (BT) model. Further, by leveraging features relevant to preference annotation, we capture uncertainty in judgments and model preference flipping patterns. In practice, we design a simple yet efficient iterative optimization algorithm compatible with the original RLHF and DPO algorithms. In our experiments, we investigate the instance-dependent preference flipping model under multiple circumstances for evaluation of our proposed method, as well as other baseline methods.