# Research Papers — 2025-12 · bundle page 65 of 175 > Full-text markdown of papers #3201–3250 of 8703 published in 2025-12, newest first. AI-readable mirrors of the full 8703-paper 2025-12 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2025-12/page/65 [← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/64.md) [Next bundle →](https://aipod.dev/papers/archive/2025-12/page/66.md) ### EverybodyDance: Bipartite Graph-Based Identity Correspondence for Multi-Character Animation Authors: Haotian Ling, Zequn Chen, Qiuying Chen, Donglin Di, Yongjia Ma, Hao Li, Chen Wei, Zhulin Tao, Xun Yang Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16360v1 Consistent pose-driven character animation has achieved remarkable progress in single-character scenarios. However, extending these advances to multi-character settings is non-trivial, especially when position swap is involved. Beyond mere scaling, the core challenge lies in enforcing correct Identity Correspondence (IC) between characters in reference and generated frames. To address this, we introduce EverybodyDance, a systematic solution targeting IC correctness in multi-character animation. EverybodyDance is built around the Identity Matching Graph (IMG), which models characters in the generated and reference frames as two node sets in a weighted complete bipartite graph. Edge weights, computed via our proposed Mask-Query Attention (MQA), quantify the affinity between each pair of characters. Our key insight is to formalize IC correctness as a graph structural metric and to optimize it during training. We also propose a series of targeted strategies tailored for multi-character animation, including identity-embedded guidance, a multi-scale matching strategy, and pre-classified sampling, which work synergistically. Finally, to evaluate IC performance, we curate the Identity Correspondence Evaluation benchmark, dedicated to multi-character IC correctness. Extensive experiments demonstrate that EverybodyDance substantially outperforms state-of-the-art baselines in both IC and visual fidelity. --- ### GMODiff: One-Step Gain Map Refinement with Diffusion Priors for HDR Reconstruction Authors: Tao Hu, Weiyu Zhou, Yanjie Tu, Peng Wu, Wei Dong, Qingsen Yan, Yanning Zhang Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16357v3 Pre-trained Latent Diffusion Models (LDMs) have recently shown strong perceptual priors for low-level vision tasks, making them a promising direction for multi-exposure High Dynamic Range (HDR) reconstruction. However, directly applying LDMs to HDR remains challenging due to: (1) limited dynamic-range representation caused by 8-bit latent compression, (2) high inference cost from multi-step denoising, and (3) content hallucination inherent to their generative nature. To address these challenges, we introduce GMODiff, a gain map-driven one-step diffusion framework for multi-exposure HDR reconstruction. Instead of reconstructing full HDR content, we reformulate HDR reconstruction as a degradation-aware Gain Map (GM) refinement problem, where the GM encodes the extended dynamic range while retaining the same bit depth as LDR images. We initialize the denoising process from an informative regression-based estimate rather than pure noise, allowing the model to generate high-quality GMs in a single denoising step. Furthermore, recognizing that regression-based models excel in content fidelity while LDMs favor perceptual quality, we leverage regression priors to guide both the denoising process and latent decoding of the LDM, suppressing hallucinations while preserving structural accuracy. Extensive experiments demonstrate that GMODiff performs favorably against several state-of-the-art methods and is 100x faster than previous LDM-based methods. Code is available at https://github.com/gbymat/GMODiff. --- ### Collaborative Edge-to-Server Inference for Vision-Language Models Authors: Soochang Song, Yongjune Kim Published: 2025-12-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.16349v2 We propose a collaborative edge-to-server inference framework for vision-language models (VLMs) that reduces communication cost while maintaining inference accuracy. In typical deployments, visual data captured at edge devices (clients) is transmitted to the server for VLM inference. However, transmitting full-resolution images incurs high communication cost. Conversely, aggressive downsizing or excessive compression to mitigate communication overhead can discard fine-grained details, leading to accuracy degradation. To overcome this limitation, we design a communication-efficient two-stage framework. In the first stage, the server performs inference on the downsized thumbnail (global image) and quantifies the min-entropy of the output tokens. If the min-entropy exceeds a predefined threshold, the server identifies a region of interest (RoI) using the VLM's internal attention and requests the edge device to send a detail-preserved local image of the RoI. The server then refines its inference by jointly leveraging the global and local images. This selective retransmission strategy ensures that only essential visual content is additionally transmitted. Experimental results consistently confirm that the proposed framework substantially reduces communication overhead while maintaining inference accuracy across diverse VQA benchmarks. --- ### AI Needs Physics More Than Physics Needs AI Authors: Peter Coveney, Roger Highfield Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16344v1 Artificial intelligence (AI) is commonly depicted as transformative. Yet, after more than a decade of hype, its measurable impact remains modest outside a few high-profile scientific and commercial successes. The 2024 Nobel Prizes in Chemistry and Physics recognized AI's potential, but broader assessments indicate the impact to date is often more promotional than technical. We argue that while current AI may influence physics, physics has significantly more to offer this generation of AI. Current architectures - large language models, reasoning models, and agentic AI - can depend on trillions of meaningless parameters, suffer from distributional bias, lack uncertainty quantification, provide no mechanistic insights, and fail to capture even elementary scientific laws. We review critiques of these limits, highlight opportunities in quantum AI and analogue computing, and lay down a roadmap for the adoption of 'Big AI': a synthesis of theory-based rigour with the flexibility of machine learning. --- ### Pretrained battery transformer (PBT): A foundation model for battery life prediction Authors: Ruifeng Tan, Weixiang Hong, Jia Li, Jiaqiang Huang, Tong-Yi Zhang Published: 2025-12-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.16334v7 Early prediction of battery cycle life is essential for improving battery design, manufacturing and deployment. However, despite encouraging progress with machine learning, battery life prediction remains constrained by scarce data and pronounced heterogeneity across battery chemistries, specifications, formation protocols and operating conditions. Although transfer learning has been widely explored to alleviate these challenges, its effectiveness is limited by the absence of a foundation model that can integrate heterogeneous battery life data and provide broadly useful knowledge for target-scenario specialization. Here we introduce the pretrained battery transformer (PBT), an integrated foundation model comprising a general PBT and specialized PBT models for individual target scenarios. At its core, battery-knowledge-encoded mixture-of-experts layers enable the general PBT to consolidate shared cycling-pattern-lifetime relationships from 13 heterogeneous lithium-ion battery datasets while preserving specialization across distinct aging regimes. The resulting general PBT provides a shared knowledge and parameter foundation from which specialized PBT models are constructed using limited labelled data to capture target-specific degradation behavior. Across 15 downstream datasets covering 977 batteries and 532 aging conditions from lithium-ion, sodium-ion and zinc-ion batteries, the specialized PBT models achieve state-of-the-art performance, outperforming the strongest comparator by 24.8% on average and by up to 73.9%. This study establishes, to our knowledge, the first foundation model for battery life prediction and points towards a shift from isolated, scenario-specific modelling to a reusable knowledge foundation for data-efficient specialization, with broader implications for sustainable-energy prediction problems constrained by scarce and heterogeneous data. --- ### QUIDS: Quality-informed Incentive-driven Multi-agent Dispatching System for Mobile Crowdsensing Authors: Nan Zhou, Zuxin Li, Fanhang Man, Xuecheng Chen, Susu Xu, Fan Dang, Chaopeng Hong, Yunhao Liu, Xiao-Ping Zhang, Xinlei Chen Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16325v1 This paper addresses the challenge of achieving optimal Quality of Information (QoI) in non-dedicated vehicular mobile crowdsensing (NVMCS) systems. The key obstacles are the interrelated issues of sensing coverage, sensing reliability, and the dynamic participation of vehicles. To tackle these, we propose QUIDS, a QUality-informed Incentive-driven multi-agent Dispatching System, which ensures high sensing coverage and reliability under budget constraints. QUIDS introduces a novel metric, Aggregated Sensing Quality (ASQ), to quantitatively capture QoI by integrating both coverage and reliability. We also develop a Mutually Assisted Belief-aware Vehicle Dispatching algorithm that estimates sensing reliability and allocates incentives under uncertainty, further improving ASQ. Evaluation using real-world data from a metropolitan NVMCS deployment shows QUIDS improves ASQ by 38% over non-dispatching scenarios and by 10% over state-of-the-art methods. It also reduces reconstruction map errors by 39-74% across algorithms. By jointly optimizing coverage and reliability via a quality-informed incentive mechanism, QUIDS enables low-cost, high-quality urban monitoring without dedicated infrastructure, applicable to smart-city scenarios like traffic and environmental sensing. --- ### Hacking Neural Evaluation Metrics with Single Hub Text Authors: Hiroyuki Deguchi, Katsuki Chousa, Yusuke Sakai Published: 2025-12-18 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.16323v2 Strongly human-correlated evaluation metrics serve as an essential compass for the development and improvement of generation models and must be highly reliable and robust. Recent embedding-based neural text evaluation metrics, such as COMET for translation tasks, are widely used in both research and development fields. However, there is no guarantee that they yield reliable evaluation results due to the black-box nature of neural networks. To raise concerns about the reliability and safety of such metrics, we propose a method for finding a single adversarial text in the discrete space that is consistently evaluated as high-quality, regardless of the test cases, to identify the vulnerabilities in evaluation metrics. The single hub text found with our method achieved 79.1 COMET% and 67.8 COMET% in the WMT'24 English-to-Japanese (En--Ja) and English-to-German (En--De) translation tasks, respectively, outperforming translations generated individually for each source sentence by using M2M100, a general translation model. Furthermore, we also confirmed that the hub text found with our method generalizes across multiple language pairs such as Ja--En and De--En. --- ### Design and Evaluation of Cost-Aware PoQ for Decentralized LLM Inference Authors: Arther Tian, Alex Ding, Frank Chen, Alan Wu, Aaron Chan, Bruce Zhang Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16317v1 Decentralized large language model (LLM) inference promises transparent and censorship resistant access to advanced AI, yet existing verification approaches struggle to scale to modern models. Proof of Quality (PoQ) replaces cryptographic verification of computation with consensus over output quality, but the original formulation ignores heterogeneous computational costs across inference and evaluator nodes. This paper introduces a cost-aware PoQ framework that integrates explicit efficiency measurements into the reward mechanism for both types of nodes. The design combines ground truth token level F1, lightweight learned evaluators, and GPT based judgments within a unified evaluation pipeline, and adopts a linear reward function that balances normalized quality and cost. Experiments on extractive question answering and abstractive summarization use five instruction tuned LLMs ranging from TinyLlama-1.1B to Llama-3.2-3B and three evaluation models spanning cross encoder and bi encoder architectures. Results show that a semantic textual similarity bi encoder achieves much higher correlation with both ground truth and GPT scores than cross encoders, indicating that evaluator architecture is a critical design choice for PoQ. Quality-cost analysis further reveals that the largest models in the pool are also the most efficient in terms of quality per unit latency. Monte Carlo simulations over 5\,000 PoQ rounds demonstrate that the cost-aware reward scheme consistently assigns higher average rewards to high quality low cost inference models and to efficient evaluators, while penalizing slow low quality nodes. These findings suggest that cost-aware PoQ provides a practical foundation for economically sustainable decentralized LLM inference. --- ### Ridge Estimation-Based Vision and Laser Ranging Fusion Localization Method for UAVs Authors: Huayu Huang, Chen Chen, Banglei Guan, Ze Tan, Yang Shang, Zhang Li, Qifeng Yu Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16314v1 Tracking and measuring targets using a variety of sensors mounted on UAVs is an effective means to quickly and accurately locate the target. This paper proposes a fusion localization method based on ridge estimation, combining the advantages of rich scene information from sequential imagery with the high precision of laser ranging to enhance localization accuracy. Under limited conditions such as long distances, small intersection angles, and large inclination angles, the column vectors of the design matrix have serious multicollinearity when using the least squares estimation algorithm. The multicollinearity will lead to ill-conditioned problems, resulting in significant instability and low robustness. Ridge estimation is introduced to mitigate the serious multicollinearity under the condition of limited observation. Experimental results demonstrate that our method achieves higher localization accuracy compared to ground localization algorithms based on single information. Moreover, the introduction of ridge estimation effectively enhances the robustness, particularly under limited observation conditions. --- ### LaverNet: Lightweight All-in-one Video Restoration via Selective Propagation Authors: Haiyu Zhao, Yiwen Shan, Yuanbiao Gou, Xi Peng Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16313v1 Recent studies have explored all-in-one video restoration, which handles multiple degradations with a unified model. However, these approaches still face two challenges when dealing with time-varying degradations. First, the degradation can dominate temporal modeling, confusing the model to focus on artifacts rather than the video content. Second, current methods typically rely on large models to handle all-in-one restoration, concealing those underlying difficulties. To address these challenges, we propose a lightweight all-in-one video restoration network, LaverNet, with only 362K parameters. To mitigate the impact of degradations on temporal modeling, we introduce a novel propagation mechanism that selectively transmits only degradation-agnostic features across frames. Through LaverNet, we demonstrate that strong all-in-one restoration can be achieved with a compact network. Despite its small size, less than 1\% of the parameters of existing models, LaverNet achieves comparable, even superior performance across benchmarks. --- ### Agent Tools Orchestration Leaks More: Dataset, Benchmark, and Mitigation Authors: Yuxuan Qiao, Dongqin Liu, Hongchang Yang, Wei Zhou, Songlin Hu Published: 2025-12-18 Categories: cs.CR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.16310v4 LLM agents can combine individually non-revealing tool returns and disclose a sensitive conclusion, creating Tools Orchestration Privacy Risk (TOP-R). We formalize TOP-R through three conditions: conclusion sensitivity, single-source non-inferability, and compositional inferability. We introduce Library-Grounded Reverse-Inference Seed Expansion (LRSE), a four-library reverse-construction pipeline, and use it to build TOP-Bench, a 1,000-instance benchmark evaluated under a controlled two-stage tool-use protocol. Across six LLM agents, average task completion, leakage, and H-score are 98.0 percent, 88.6 percent, and 20.4. With native reasoning enabled, four models average 81.4 percent final-response leakage and 82.4 percent reasoning-trace leakage. With reasoning disabled, three prompt-only safeguards improve H-score by an average of about 3.4 points on TOP-Bench. We further propose TOP-Align, an SFT+DPO method for learning safer task-completion boundaries. On a separate post-training evaluation set, TOP-Align improves H-score by 16.2 points over the base model, versus a 5.0-point average gain from prompt-only mitigation on the same set. These results show that TOP-R requires defenses beyond prompting alone. Dataset and code are available at https://github.com/1Ponder/TOP-R. --- ### Beyond the Benchmark: Innovative Defenses Against Prompt Injection Attacks Authors: Safwan Shaheer, G. M. Refatul Islam, Mohammad Rafid Hamid, Tahsin Zaman Jilan Published: 2025-12-18 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2512.16307v1 In this fast-evolving area of LLMs, our paper discusses the significant security risk presented by prompt injection attacks. It focuses on small open-sourced models, specifically the LLaMA family of models. We introduce novel defense mechanisms capable of generating automatic defenses and systematically evaluate said generated defenses against a comprehensive set of benchmarked attacks. Thus, we empirically demonstrated the improvement proposed by our approach in mitigating goal-hijacking vulnerabilities in LLMs. Our work recognizes the increasing relevance of small open-sourced LLMs and their potential for broad deployments on edge devices, aligning with future trends in LLM applications. We contribute to the greater ecosystem of open-source LLMs and their security in the following: (1) assessing present prompt-based defenses against the latest attacks, (2) introducing a new framework using a seed defense (Chain Of Thoughts) to refine the defense prompts iteratively, and (3) showing significant improvements in detecting goal hijacking attacks. Out strategies significantly reduce the success rates of the attacks and false detection rates while at the same time effectively detecting goal-hijacking capabilities, paving the way for more secure and efficient deployments of small and open-source LLMs in resource-constrained environments. --- ### Can Transformers overcome the lack of data in the simulation of history-dependent flows? Authors: P. Urdeitx, I. Alfaro, D. Gonzalez, F. Chinesta, E. Cueto Published: 2025-12-18 Categories: physics.flu-dyn, cs.LG arXiv: https://arxiv.org/abs/2512.16305v1 It is well known that the lack of information about certain variables necessary for the description of a dynamical system leads to the introduction of historical dependence (lack of Markovian character of the model) and noise. Traditionally, scientists have made up for these shortcomings by designing phenomenological variables that take into account this historical dependence (typically, conformational tensors in fluids). Often, these phenomenological variables are not easily measurable experimentally. In this work, we study to what extent Transformer architectures are able to cope with the lack of experimental data on these variables. The methodology is evaluated on three benchmark problems: a cylinder flow with no history dependence, a viscoelastic Couette flow modeled via the Oldroyd-B formalism, and a non-linear polymeric fluid described by the FENE model. Our results show that the Transformer outperforms a thermodynamically consistent, structure-preserving neural network with metriplectic bias in systems with missing experimental data, providing lower errors even in low-dimensional latent spaces. In contrast, for systems whose state variables can be fully known, the metriplectic model achieves superior performance. --- ### PixelArena: A benchmark for Pixel-Precision Visual Intelligence Authors: Feng Liang, Sizhe Cheng, Chenqi Yi, Yong Wang Published: 2025-12-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.16303v2 Omni-modal models that have multimodal input and output are emerging. However, benchmarking their multimodal generation, especially in image generation, is challenging due to the subtleties of human preferences and model biases. Many image generation benchmarks focus on aesthetics instead of the fine-grained generation capabilities of these models, failing to evaluate their visual intelligence with objective metrics. In PixelArena, we propose using semantic segmentation tasks to objectively examine their fine-grained generative intelligence with pixel precision. With our benchmark and experiments, we find the latest Gemini 3 Pro Image has emergent image generation capabilities that generate semantic masks with high fidelity under zero-shot settings, showcasing visual intelligence unseen before and true generalization in new image generation tasks. We further investigate its results, compare them qualitatively and quantitatively with those of other models, and present failure cases. The findings not only signal exciting progress in the field but also provide insights into future research related to dataset development, omni-modal model development, and the design of metrics. --- ### Adaptation of Agentic AI: A Survey of Post-Training, Memory, and Skills Authors: Pengcheng Jiang, Jiacheng Lin, Zhiyi Shi, Zifeng Wang, Luxi He, Yichen Wu, Ming Zhong, Peiyang Song, Qizheng Zhang, Heng Wang, Xueqiang Xu, Hanwen Xu et al. Published: 2025-12-18 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.16301v3 Large language model (LLM) agents are moving beyond prompting alone. ChatGPT marked the rise of general-purpose LLM assistants, DeepSeek showed that on-policy reinforcement learning with verifiable rewards can improve reasoning and tool use, and OpenClaw highlights a newer direction in which agents accumulate persistent memory and reusable skills. Yet the research landscape remains fragmented across post-training, retrieval, memory, and skill systems. This survey studies these developments under a single notion of \emph{adaptation}: improving an agent, its tools, or their interaction after pretraining. We organize the field with a four-paradigm framework spanning agent adaptation and tool adaptation. On the agent side, A1 (tool-execution-signaled) and A2 (agent-output-signaled) improve the agent itself through supervised fine-tuning, preference optimization, and reinforcement learning with verifiable rewards. On the tool side, T1 (agent-agnostic) provides reusable pre-trained modules any agent can call, while T2 (agent-supervised) uses the agent's outputs to train memory systems, skill libraries, or lightweight subagents. Using this framework, we review post-training methods, adaptive memory architectures, and agent skills; compare their trade-offs in cost, flexibility, and generalization; and summarize evaluation practices across deep research, software development, computer use, and drug discovery. We conclude by outlining open problems in agent-tool co-adaptation, continual learning, safety, and efficient deployment. --- ### Code-in-the-Loop Forensics: Agentic Tool Use for Image Forgery Detection Authors: Fanrui Zhang, Qiang Zhang, Sizhuo Zhou, Jianwen Sun, Chuanhao Li, Jiaxin Ai, Yukang Feng, Yujie Zhang, Wenjie Li, Zizhen Li, Yifan Chang, Jiawei Liu et al. Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16300v3 Existing image forgery detection (IFD) methods either exploit low-level, semantics-agnostic artifacts or rely on multimodal large language models (MLLMs) with high-level semantic knowledge. Although naturally complementary, these two information streams are highly heterogeneous in both paradigm and reasoning, making it difficult for existing methods to unify them or effectively model their cross-level interactions. To address this gap, we propose ForenAgent, a multi-round interactive IFD framework that enables MLLMs to autonomously generate, execute, and iteratively refine Python-based low-level tools around the detection objective, thereby achieving more flexible and interpretable forgery analysis. ForenAgent follows a two-stage training pipeline combining Cold Start and Reinforcement Fine-Tuning to enhance its tool interaction capability and reasoning adaptability progressively. Inspired by human reasoning, we design a dynamic reasoning loop comprising global perception, local focusing, iterative probing, and holistic adjudication, and instantiate it as both a data-sampling strategy and a task-aligned process reward. For systematic training and evaluation, we construct FABench, a heterogeneous, high-quality agent-forensics dataset comprising 100k images and approximately 200k agent-interaction question-answer pairs. Experiments show that ForenAgent exhibits emergent tool-use competence and reflective reasoning on challenging IFD tasks when assisted by low-level tools, charting a promising route toward general-purpose IFD. The code will be released after the review process is completed. --- ### Feature-Selective Representation Misdirection for Machine Unlearning Authors: Taozhao Chen, Linghan Huang, Kim-Kwang Raymond Choo, Huaming Chen Published: 2025-12-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.16297v1 As large language models (LLMs) are increasingly adopted in safety-critical and regulated sectors, the retention of sensitive or prohibited knowledge introduces escalating risks, ranging from privacy leakage to regulatory non-compliance to to potential misuse, and so on. Recent studies suggest that machine unlearning can help ensure deployed models comply with evolving legal, safety, and governance requirements. However, current unlearning techniques assume clean separation between forget and retain datasets, which is challenging in operational settings characterized by highly entangled distributions. In such scenarios, perturbation-based methods often degrade general model utility or fail to ensure safety. To address this, we propose Selective Representation Misdirection for Unlearning (SRMU), a novel principled activation-editing framework that enforces feature-aware and directionally controlled perturbations. Unlike indiscriminate model weights perturbations, SRMU employs a structured misdirection vector with an activation importance map. The goal is to allow SRMU selectively suppresses harmful representations while preserving the utility on benign ones. Experiments are conducted on the widely used WMDP benchmark across low- and high-entanglement configurations. Empirical results reveal that SRMU delivers state-of-the-art unlearning performance with minimal utility losses, and remains effective under 20-30\% overlap where existing baselines collapse. SRMU provides a robust foundation for safety-driven model governance, privacy compliance, and controlled knowledge removal in the emerging LLM-based applications. We release the replication package at https://figshare.com/s/d5931192a8824de26aff. --- ### OS-Oracle: A Comprehensive Framework for Cross-Platform GUI Critic Models Authors: Zhenyu Wu, Jingjing Xie, Zehao Li, Bowen Yang, Qiushi Sun, Zhaoyang Liu, Zhoumianze Liu, Yu Qiao, Xiangyu Yue, Zun Wang, Zichen Ding Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16295v1 With VLM-powered computer-using agents (CUAs) becoming increasingly capable at graphical user interface (GUI) navigation and manipulation, reliable step-level decision-making has emerged as a key bottleneck for real-world deployment. In long-horizon workflows, errors accumulate quickly and irreversible actions can cause unintended consequences, motivating critic models that assess each action before execution. While critic models offer a promising solution, their effectiveness is hindered by the lack of diverse, high-quality GUI feedback data and public critic benchmarks for step-level evaluation in computer use. To bridge these gaps, we introduce OS-Oracle that makes three core contributions: (1) a scalable data pipeline for synthesizing cross-platform GUI critic data; (2) a two-stage training paradigm combining supervised fine-tuning (SFT) and consistency-preserving group relative policy optimization (CP-GRPO); (3) OS-Critic Bench, a holistic benchmark for evaluating critic model performance across Mobile, Web, and Desktop platforms. Leveraging this framework, we curate a high-quality dataset containing 310k critic samples. The resulting critic model, OS-Oracle-7B, achieves state-of-the-art performance among open-source VLMs on OS-Critic Bench, and surpasses proprietary models on the mobile domain. Furthermore, when serving as a pre-critic, OS-Oracle-7B improves the performance of native GUI agents such as UI-TARS-1.5-7B in OSWorld and AndroidWorld environments. The code is open-sourced at https://github.com/numbmelon/OS-Oracle. --- ### MARC: Multi-Label Adaptive Retrieval Contrastive Loss for Remote Sensing Images Authors: Amna Amir, Erchan Aptoula Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16294v2 Semantic overlap among land-cover categories, highly imbalanced label distributions, and complex inter-class co-occurrence patterns constitute significant challenges for multi-label remote-sensing image retrieval. In this article, Multi-Label Adaptive Contrastive Learning (MACL) is introduced as an extension of contrastive learning to address them. It integrates label-aware sampling, frequency-sensitive weighting, and dynamic-temperature scaling to achieve balanced representation learning across both common and rare categories. Extensive experiments on three benchmark datasets (DLRSD, ML-AID, and WHDLD), show that MACL consistently outperforms contrastive-loss based baselines, effectively mitigating semantic imbalance and delivering more reliable retrieval performance in large-scale remote-sensing archives. Code, pretrained models, and evaluation scripts will be released at https://github.com/Amna-128/MARC upon acceptance. --- ### In-Context Probing for Membership Inference in Fine-Tuned Language Models Authors: Zhexi Lu, Hongliang Chi, Nathalie Baracaldo, Swanand Ravindra Kadhe, Yuseok Jeon, Lei Yu Published: 2025-12-18 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2512.16292v2 Membership inference attacks (MIAs) pose a critical privacy threat to fine-tuned large language models (LLMs), especially when models are adapted to domain-specific tasks using sensitive data. While prior black-box MIA techniques rely on confidence scores or token likelihoods, these signals are often entangled with a sample's intrinsic properties - such as content difficulty or rarity - leading to poor generalization and low signal-to-noise ratios. In this paper, we propose ICP-MIA, a novel MIA framework grounded in the theory of training dynamics, particularly the phenomenon of diminishing returns during optimization. We introduce the Optimization Gap as a fundamental signal of membership: at convergence, member samples exhibit minimal remaining loss-reduction potential, while non-members retain significant potential for further optimization. To estimate this gap in a black-box setting, we propose In-Context Probing (ICP), a training-free method that simulates fine-tuning-like behavior via strategically constructed input contexts. We propose two probing strategies: reference-data-based (using semantically similar public samples) and self-perturbation (via masking or generation). Experiments on three tasks and multiple LLMs show that ICP-MIA significantly outperforms prior black-box MIAs, particularly at low false positive rates. We further analyze how reference data alignment, model type, PEFT configurations, and training schedules affect attack effectiveness. Our findings establish ICP-MIA as a practical and theoretically grounded framework for auditing privacy risks in deployed LLMs. --- ### Evaluating OpenAI GPT Models for Translation of Endangered Uralic Languages: A Comparison of Reasoning and Non-Reasoning Architectures Authors: Yehor Tereshchenko, Mika Hämäläinen, Svitlana Myroniuk Published: 2025-12-18 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.16287v1 The evaluation of Large Language Models (LLMs) for translation tasks has primarily focused on high-resource languages, leaving a significant gap in understanding their performance on low-resource and endangered languages. This study presents a comprehensive comparison of OpenAI's GPT models, specifically examining the differences between reasoning and non-reasoning architectures for translating between Finnish and four low-resource Uralic languages: Komi-Zyrian, Moksha, Erzya, and Udmurt. Using a parallel corpus of literary texts, we evaluate model willingness to attempt translation through refusal rate analysis across different model architectures. Our findings reveal significant performance variations between reasoning and non-reasoning models, with reasoning models showing 16 percentage points lower refusal rates. The results provide valuable insights for researchers and practitioners working with Uralic languages and contribute to the broader understanding of reasoning model capabilities for endangered language preservation. --- ### CALM: A CKA-Guided Adaptive Layer-Wise Modularization Framework for LLM Quantization Authors: Jinhao Zhang, Yunquan Zhang, Daning Chen, JunSun, Zicheng Yan Published: 2025-12-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.16282v2 Current mainstream post-training quantization methods for large language models typically apply a uniform quantization strategy across all network layers, overlooking the substantial differences in algorithmic suitability among layers. To address this limitation, we propose CALM (A CKA-guided Adaptive Layer-wise Modularization)a fine-tuning-free, plug-and-play framework for algorithmic heterogeneous quantization. CALM independently evaluates multiple PTQ algorithms on each layer and employs Linear Centered Kernel Alignment (CKA) as a metric to automatically select the optimal quantization strategy per layer. The individually optimized strategies are then integrated to construct a hybrid quantized model. Experiments demonstrate that our approach consistently outperforms both uniform quantization baselines and state-of-the-art mixed-precision methods across mainstream LLMsincluding LLaMA and Qwenin terms of perplexity (PPL) and downstream task performance. --- ### Love, Lies, and Language Models: Investigating AI's Role in Romance-Baiting Scams Authors: Gilad Gressel, Rahul Pankajakshan, Shir Rozenfeld, Ling Li, Ivan Franceschini, Krishnashree Achuthan, Yisroel Mirsky Published: 2025-12-18 Categories: cs.CR, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2512.16280v3 Romance-baiting scams have become a major source of financial and emotional harm worldwide. These operations are run by organized crime syndicates that traffic thousands of people into forced labor, requiring them to build emotional intimacy with victims over weeks of text conversations before pressuring them into fraudulent cryptocurrency investments. Because the scams are inherently text-based, they raise urgent questions about the role of Large Language Models (LLMs) in both current and future automation. We investigate this intersection by interviewing 145 insiders and 5 scam victims, performing a blinded long-term conversation study comparing LLM scam agents to human operators, and executing an evaluation of commercial safety filters. Our findings show that LLMs are already widely deployed within scam organizations, with 87% of scam labor consisting of systematized conversational tasks readily susceptible to automation. In a week-long study, an LLM agent not only elicited greater trust from study participants (p=0.007) but also achieved higher compliance with requests than human operators (46% vs. 18% for humans). Meanwhile, popular safety filters detected 0.0% of romance baiting dialogues. Together, these results suggest that romance-baiting scams may be amenable to full-scale LLM automation, while existing defenses remain inadequate to prevent their expansion. --- ### QuadSentinel: Sequent Safety for Machine-Checkable Control in Multi-agent Systems Authors: Yiliu Yang, Yilei Jiang, Qunzhong Wang, Yingshui Tan, Xiaoyong Zhu, Sherman S. M. Chow, Bo Zheng, Xiangyu Yue Published: 2025-12-18 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.16279v1 Safety risks arise as large language model-based agents solve complex tasks with tools, multi-step plans, and inter-agent messages. However, deployer-written policies in natural language are ambiguous and context dependent, so they map poorly to machine-checkable rules, and runtime enforcement is unreliable. Expressing safety policies as sequents, we propose \textsc{QuadSentinel}, a four-agent guard (state tracker, policy verifier, threat watcher, and referee) that compiles these policies into machine-checkable rules built from predicates over observable state and enforces them online. Referee logic plus an efficient top-$k$ predicate updater keeps costs low by prioritizing checks and resolving conflicts hierarchically. Measured on ST-WebAgentBench (ICML CUA~'25) and AgentHarm (ICLR~'25), \textsc{QuadSentinel} improves guardrail accuracy and rule recall while reducing false positives. Against single-agent baselines such as ShieldAgent (ICML~'25), it yields better overall safety control. Near-term deployments can adopt this pattern without modifying core agents by keeping policies separate and machine-checkable. Our code will be made publicly available at https://github.com/yyiliu/QuadSentinel. --- ### Sharpness-aware Second-order Latent Factor Model for High-dimensional and Incomplete Data Authors: Jialiang Wang, Xueyan Bao, Hao Wu Published: 2025-12-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.16277v1 Second-order Latent Factor (SLF) model, a class of low-rank representation learning methods, has proven effective at extracting node-to-node interaction patterns from High-dimensional and Incomplete (HDI) data. However, its optimization is notoriously difficult due to its bilinear and non-convex nature. Sharpness-aware Minimization (SAM) has recently proposed to find flat local minima when minimizing non-convex objectives, thereby improving the generalization of representation-learning models. To address this challenge, we propose a Sharpness-aware SLF (SSLF) model. SSLF embodies two key ideas: (1) acquiring second-order information via Hessian-vector products; and (2) injecting a sharpness term into the curvature (Hessian) through the designed Hessian-vector products. Experiments on multiple industrial datasets demonstrate that the proposed model consistently outperforms state-of-the-art baselines. --- ### GFLAN: Generative Functional Layouts Authors: Mohamed Abouagour, Eleftherios Garyfallidis Published: 2025-12-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.16275v2 Automated floor plan generation lies at the intersection of combinatorial search, geometric constraint satisfaction, and functional design requirements -- a confluence that has historically resisted a unified computational treatment. While recent deep learning approaches have improved the state of the art, they often struggle to capture architectural reasoning: the precedence of topological relationships over geometric instantiation, the propagation of functional constraints through adjacency networks, and the emergence of circulation patterns from local connectivity decisions. To address these fundamental challenges, this paper introduces GFLAN, a generative framework that restructures floor plan synthesis through explicit factorization into topological planning and geometric realization. Given a single exterior boundary and a front-door location, our approach departs from direct pixel-to-pixel or wall-tracing generation in favor of a principled two-stage decomposition. Stage A employs a specialized convolutional architecture with dual encoders -- separating invariant spatial context from evolving layout state -- to sequentially allocate room centroids within the building envelope via discrete probability maps over feasible placements. Stage B constructs a heterogeneous graph linking room nodes to boundary vertices, then applies a Transformer-augmented graph neural network (GNN) that jointly regresses room boundaries. --- ### Beyond Blind Spots: Analytic Hints for Mitigating LLM-Based Evaluation Pitfalls Authors: Ora Nova Fandina, Eitan Farchi, Shmulik Froimovich, Raviv Gal, Wesam Ibraheem, Rami Katan, Alice Podolsky Published: 2025-12-18 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2512.16272v2 Large Language Models are increasingly deployed as judges (LaaJ) in code generation pipelines. While attractive for scalability, LaaJs tend to overlook domain specific issues raising concerns about their reliability in critical evaluation tasks. To better understand these limitations in practice, we examine LaaJ behavior in a concrete industrial use case: legacy code modernization via COBOL code generation. In this setting, we find that even production deployed LaaJs can miss domain critical errors, revealing consistent blind spots in their evaluation capabilities. To better understand these blind spots, we analyze generated COBOL programs and associated LaaJs judgments, drawing on expert knowledge to construct a preliminary taxonomy. Based on this taxonomy, we develop a lightweight analytic checker tool that flags over 30 domain specific issues observed in practice. We use its outputs as analytic hints, dynamically injecting them into the judges prompt to encourage LaaJ to revisit aspects it may have overlooked. Experiments on a test set of 100 programs using four production level LaaJs show that LaaJ alone detects only about 45-63% of the errors present in the code (in all judges we tested), while the analytic checker alone lacks explanatory depth. When combined, the LaaJ+Hints configuration achieves up to 74% coverage (for the best performing judge and injection prompt) and produces qualitatively richer, more accurate explanations, demonstrating that analytic-LLM hybrids can substantially enhance evaluation reliability in deployed pipelines. We release the dataset and all used prompts. --- ### Domain-Agnostic Causal-Aware Audio Transformer for Infant Cry Classification Authors: Geofrey Owino, Bernard Shibwabo Kasamani, Ahmed M. Abdelmoniem, Edem Wornyo Published: 2025-12-18 Categories: cs.SD, cs.AI arXiv: https://arxiv.org/abs/2512.16271v1 Accurate and interpretable classification of infant cry paralinguistics is essential for early detection of neonatal distress and clinical decision support. However, many existing deep learning methods rely on correlation-driven acoustic representations, which makes them vulnerable to noise, spurious cues, and domain shifts across recording environments. We propose DACH-TIC, a Domain-Agnostic Causal-Aware Hierarchical Audio Transformer for robust infant cry classification. The model integrates causal attention, hierarchical representation learning, multi-task supervision, and adversarial domain generalization within a unified framework. DACH-TIC employs a structured transformer backbone with local token-level and global semantic encoders, augmented by causal attention masking and controlled perturbation training to approximate counterfactual acoustic variations. A domain-adversarial objective promotes environment-invariant representations, while multi-task learning jointly optimizes cry type recognition, distress intensity estimation, and causal relevance prediction. The model is evaluated on the Baby Chillanto and Donate-a-Cry datasets, with ESC-50 environmental noise overlays for domain augmentation. Experimental results show that DACH-TIC outperforms state-of-the-art baselines, including HTS-AT and SE-ResNet Transformer, achieving improvements of 2.6 percent in accuracy and 2.2 points in macro-F1 score, alongside enhanced causal fidelity. The model generalizes effectively to unseen acoustic environments, with a domain performance gap of only 2.4 percent, demonstrating its suitability for real-world neonatal acoustic monitoring systems. --- ### TextEditBench: Evaluating Reasoning-aware Text Editing Beyond Rendering Authors: Rui Gui, Yang Wan, Haochen Han, Dongxing Mao, Fangming Liu, Min Li, Alex Jinpeng Wang Published: 2025-12-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.16270v1 Text rendering has recently emerged as one of the most challenging frontiers in visual generation, drawing significant attention from large-scale diffusion and multimodal models. However, text editing within images remains largely unexplored, as it requires generating legible characters while preserving semantic, geometric, and contextual coherence. To fill this gap, we introduce TextEditBench, a comprehensive evaluation benchmark that explicitly focuses on text-centric regions in images. Beyond basic pixel manipulations, our benchmark emphasizes reasoning-intensive editing scenarios that require models to understand physical plausibility, linguistic meaning, and cross-modal dependencies. We further propose a novel evaluation dimension, Semantic Expectation (SE), which measures reasoning ability of model to maintain semantic consistency, contextual coherence, and cross-modal alignment during text editing. Extensive experiments on state-of-the-art editing systems reveal that while current models can follow simple textual instructions, they still struggle with context-dependent reasoning, physical consistency, and layout-aware integration. By focusing evaluation on this long-overlooked yet fundamental capability, TextEditBench establishes a new testing ground for advancing text-guided image editing and reasoning in multimodal generation. --- ### Pixel Super-Resolved Fluorescence Lifetime Imaging Using Deep Learning Authors: Paloma Casteleiro Costa, Parnian Ghapandar Kashani, Xuhui Liu, Alexander Chen, Ary Portes, Julien Bec, Laura Marcu, Aydogan Ozcan Published: 2025-12-18 Categories: cs.CV, cs.LG, physics.med-ph, physics.optics arXiv: https://arxiv.org/abs/2512.16266v1 Fluorescence lifetime imaging microscopy (FLIM) is a powerful quantitative technique that provides metabolic and molecular contrast, offering strong translational potential for label-free, real-time diagnostics. However, its clinical adoption remains limited by long pixel dwell times and low signal-to-noise ratio (SNR), which impose a stricter resolution-speed trade-off than conventional optical imaging approaches. Here, we introduce FLIM_PSR_k, a deep learning-based multi-channel pixel super-resolution (PSR) framework that reconstructs high-resolution FLIM images from data acquired with up to a 5-fold increased pixel size. The model is trained using the conditional generative adversarial network (cGAN) framework, which, compared to diffusion model-based alternatives, delivers a more robust PSR reconstruction with substantially shorter inference times, a crucial advantage for practical deployment. FLIM_PSR_k not only enables faster image acquisition but can also alleviate SNR limitations in autofluorescence-based FLIM. Blind testing on held-out patient-derived tumor tissue samples demonstrates that FLIM_PSR_k reliably achieves a super-resolution factor of k = 5, resulting in a 25-fold increase in the space-bandwidth product of the output images and revealing fine architectural features lost in lower-resolution inputs, with statistically significant improvements across various image quality metrics. By increasing FLIM's effective spatial resolution, FLIM_PSR_k advances lifetime imaging toward faster, higher-resolution, and hardware-flexible implementations compatible with low-numerical-aperture and miniaturized platforms, better positioning FLIM for translational applications. --- ### Privacy-Aware Sharing of Raw Spatial Sensor Data for Cooperative Perception Authors: Bangya Liu, Chengpo Yan, Chenghao Jiang, Suman Banerjee, Akarsh Prabhakara Published: 2025-12-18 Categories: cs.NI, cs.CV, cs.RO arXiv: https://arxiv.org/abs/2512.16265v1 Cooperative perception between vehicles is poised to offer robust and reliable scene understanding. Recently, we are witnessing experimental systems research building testbeds that share raw spatial sensor data for cooperative perception. While there has been a marked improvement in accuracies and is the natural way forward, we take a moment to consider the problems with such an approach for eventual adoption by automakers. In this paper, we first argue that new forms of privacy concerns arise and discourage stakeholders to share raw sensor data. Next, we present SHARP, a research framework to minimize privacy leakage and drive stakeholders towards the ambitious goal of raw data based cooperative perception. Finally, we discuss open questions for networked systems, mobile computing, perception researchers, industry and government in realizing our proposed framework. --- ### Learning to Wait: Synchronizing Agents with the Physical World Authors: Yifei She, Ping Zhang, He Liu, Yanmin Jia, Yang Jing, Zijun Liu, Peng Sun, Xiangbin Li, Xiaohe Hu Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16262v1 Real-world agentic tasks, unlike synchronous Markov Decision Processes (MDPs), often involve non-blocking actions with variable latencies, creating a fundamental \textit{Temporal Gap} between action initiation and completion. Existing environment-side solutions, such as blocking wrappers or frequent polling, either limit scalability or dilute the agent's context window with redundant observations. In this work, we propose an \textbf{Agent-side Approach} that empowers Large Language Models (LLMs) to actively align their \textit{Cognitive Timeline} with the physical world. By extending the Code-as-Action paradigm to the temporal domain, agents utilize semantic priors and In-Context Learning (ICL) to predict precise waiting durations (\texttt{time.sleep(t)}), effectively synchronizing with asynchronous environment without exhaustive checking. Experiments in a simulated Kubernetes cluster demonstrate that agents can precisely calibrate their internal clocks to minimize both query overhead and execution latency, validating that temporal awareness is a learnable capability essential for autonomous evolution in open-ended environments. --- ### Improving Low-Latency Learning Performance in Spiking Neural Networks via a Change-Perceptive Dendrite-Soma-Axon Neuron Authors: Zeyu Huang, Wei Meng, Quan Liu, Kun Chen, Li Ma Published: 2025-12-18 Categories: cs.NE arXiv: https://arxiv.org/abs/2512.16259v1 Spiking neurons, the fundamental information processing units of Spiking Neural Networks (SNNs), have the all-or-zero information output form that allows SNNs to be more energy-efficient compared to Artificial Neural Networks (ANNs). However, the hard reset mechanism employed in spiking neurons leads to information degradation due to its uniform handling of diverse membrane potentials. Furthermore, the utilization of overly simplified neuron models that disregard the intricate biological structures inherently impedes the network's capacity to accurately simulate the actual potential transmission process. To address these issues, we propose a dendrite-soma-axon (DSA) neuron employing the soft reset strategy, in conjunction with a potential change-based perception mechanism, culminating in the change-perceptive dendrite-soma-axon (CP-DSA) neuron. Our model contains multiple learnable parameters that expand the representation space of neurons. The change-perceptive (CP) mechanism enables our model to achieve competitive performance in short time steps utilizing the difference information of adjacent time steps. Rigorous theoretical analysis is provided to demonstrate the efficacy of the CP-DSA model and the functional characteristics of its internal parameters. Furthermore, extensive experiments conducted on various datasets substantiate the significant advantages of the CP-DSA model over state-of-the-art approaches. --- ### Interpretable Deep Learning for Stock Returns: A Consensus-Bottleneck Asset Pricing Model Authors: Changeun Kim, Younwoo Jeong, Bong-Gyu Jang Published: 2025-12-18 Categories: q-fin.PR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.16251v5 We introduce the Consensus-Bottleneck Asset Pricing Model (CB-APM), which embeds aggregate analyst consensus as a structural bottleneck, treating professional beliefs as a sufficient statistic for the market's high-dimensional information set. Unlike post-hoc explainability approaches, CB-APM achieves interpretability-by-design: the bottleneck constraint functions as an endogenous regularizer that simultaneously improves out-of-sample predictive accuracy and anchors inference to economically interpretable drivers. Portfolios sorted on CB-APM forecasts exhibit a strong monotonic return gradient, robust across macroeconomic regimes. Pricing diagnostics further reveal that the learned consensus encodes priced variation not spanned by canonical factor models, identifying belief-driven risk heterogeneity that standard linear frameworks systematically miss. --- ### AMUSE: Audio-Visual Benchmark and Alignment Framework for Agentic Multi-Speaker Understanding Authors: Sanjoy Chowdhury, Karren D. Yang, Xudong Liu, Fartash Faghri, Pavan Kumar Anasosalu Vasu, Oncel Tuzel, Dinesh Manocha, Chun-Liang Li, Raviteja Vemulapalli Published: 2025-12-18 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2512.16250v1 Recent multimodal large language models (MLLMs) such as GPT-4o and Qwen3-Omni show strong perception but struggle in multi-speaker, dialogue-centric settings that demand agentic reasoning tracking who speaks, maintaining roles, and grounding events across time. These scenarios are central to multimodal audio-video understanding, where models must jointly reason over audio and visual streams in applications such as conversational video assistants and meeting analytics. We introduce AMUSE, a benchmark designed around tasks that are inherently agentic, requiring models to decompose complex audio-visual interactions into planning, grounding, and reflection steps. It evaluates MLLMs across three modes zero-shot, guided, and agentic and six task families, including spatio-temporal speaker grounding and multimodal dialogue summarization. Across all modes, current models exhibit weak multi-speaker reasoning and inconsistent behavior under both non-agentic and agentic evaluation. Motivated by the inherently agentic nature of these tasks and recent advances in LLM agents, we propose RAFT, a data-efficient agentic alignment framework that integrates reward optimization with intrinsic multimodal self-evaluation as reward and selective parameter adaptation for data and parameter efficient updates. Using RAFT, we achieve up to 39.52\% relative improvement in accuracy on our benchmark. Together, AMUSE and RAFT provide a practical platform for examining agentic reasoning in multimodal models and improving their capabilities. --- ### Sigma-MoE-Tiny Technical Report Authors: Qingguo Hu, Zhenghao Lin, Ziyue Yang, Yucheng Ding, Xiao Liu, Yuting Jiang, Ruizhe Wang, Tianyu Chen, Zhongxin Guo, Yifan Xiong, Rui Gao, Lei Qu et al. Published: 2025-12-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2512.16248v2 Mixture-of-Experts (MoE) has emerged as a promising paradigm for foundation models due to its efficient and powerful scalability. In this work, we present Sigma-MoE-Tiny, an MoE language model that achieves the highest sparsity compared to existing open-source models. Sigma-MoE-Tiny employs fine-grained expert segmentation with up to 96 experts per layer, while activating only one expert for each token, resulting in 20B total parameters with just 0.5B activated. The major challenge introduced by such extreme sparsity lies in expert load balancing. We find that the widely-used load balancing loss tends to become ineffective in the lower layers under this setting. To address this issue, we propose a progressive sparsification schedule aiming to balance expert utilization and training stability. Sigma-MoE-Tiny is pre-trained on a diverse and high-quality corpus, followed by post-training to further unlock its capabilities. The entire training process remains remarkably stable, with no occurrence of irrecoverable loss spikes. Comprehensive evaluations reveal that, despite activating only 0.5B parameters, Sigma-MoE-Tiny achieves top-tier performance among counterparts of comparable or significantly larger scale. In addition, we provide an in-depth discussion of load balancing in highly sparse MoE models, offering insights for advancing sparsity in future MoE architectures. Project page: https://qghuxmu.github.io/Sigma-MoE-Tiny Code: https://github.com/microsoft/ltp-megatron-lm --- ### Sharpness-aware Federated Graph Learning Authors: Ruiyu Li, Peige Zhao, Guangxia Li, Pengcheng Wu, Xingyu Gao, Zhiqiang Xu Published: 2025-12-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.16247v1 One of many impediments to applying graph neural networks (GNNs) to large-scale real-world graph data is the challenge of centralized training, which requires aggregating data from different organizations, raising privacy concerns. Federated graph learning (FGL) addresses this by enabling collaborative GNN model training without sharing private data. However, a core challenge in FGL systems is the variation in local training data distributions among clients, known as the data heterogeneity problem. Most existing solutions suffer from two problems: (1) The typical optimizer based on empirical risk minimization tends to cause local models to fall into sharp valleys and weakens their generalization to out-of-distribution graph data. (2) The prevalent dimensional collapse in the learned representations of local graph data has an adverse impact on the classification capacity of the GNN model. To this end, we formulate a novel optimization objective that is aware of the sharpness (i.e., the curvature of the loss surface) of local GNN models. By minimizing the loss function and its sharpness simultaneously, we seek out model parameters in a flat region with uniformly low loss values, thus improving the generalization over heterogeneous data. By introducing a regularizer based on the correlation matrix of local representations, we relax the correlations of representations generated by individual local graph samples, so as to alleviate the dimensional collapse of the learned model. The proposed \textbf{S}harpness-aware f\textbf{E}derated gr\textbf{A}ph \textbf{L}earning (SEAL) algorithm can enhance the classification accuracy and generalization ability of local GNN models in federated graph learning. Experimental studies on several graph classification benchmarks show that SEAL consistently outperforms SOTA FGL baselines and provides gains for more participants. --- ### AlignMerge - Alignment-Preserving Large Language Model Merging via Fisher-Guided Geometric Constraints Authors: Aniruddha Roy, Jyoti Patel, Aman Chadha, Vinija Jain, Amitava Das Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16245v2 Merging large language models (LLMs) is a practical way to compose capabilities from multiple fine-tuned checkpoints without retraining. Yet standard schemes (linear weight soups, task vectors, and Fisher-weighted averaging) can preserve loss while quietly destroying alignment. We argue that merging is not a numerical trick but a geometry-constrained operation around an already-aligned anchor: fusion must be steered to respect safety geometry, not validated post hoc. We introduce AlignMerge, a geometry-aware merging framework that makes alignment an explicit invariant. In a local Fisher chart around an instruction-tuned base, we estimate an alignment subspace with projector P_A and optimize: L_AlignMerge = L_geo + lambda_align * L_align + lambda_bud * L_bud, where L_geo keeps the merge close to its experts in Fisher-Rao geometry, L_align penalizes motion along alignment-sensitive directions, and L_bud enforces a soft alignment budget. As the alignment functional we use the decoding-invariant Alignment Quality Index (AQI), a latent-space criterion that captures how cleanly aligned and misaligned behaviors separate in representation space. Across five model families (LLaMA-3 8B, Mistral 7B, Qwen 2, Phi-3.5, Gemma 2), merging safety anchors with task experts, AlignMerge improves alignment metrics (AQI, toxicity, LLM-judge alignment) while matching or exceeding the best expert on instruction-following, reasoning, and helpfulness. It also exhibits smaller alignment-subspace drift and fewer budget violations than Fisher soups, TIES, SafeMerge, and MergeAlign. These results make alignment-preserving merging a first-class design goal and suggest a path to geometry-aware composition of future foundation models. --- ### Coarse-to-Fine Open-Set Graph Node Classification with Large Language Models Authors: Xueqi Ma, Xingjun Ma, Sarah Monazam Erfani, Danilo Mandic, James Bailey Published: 2025-12-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.16244v2 Developing open-set classification methods capable of classifying in-distribution (ID) data while detecting out-of-distribution (OOD) samples is essential for deploying graph neural networks (GNNs) in open-world scenarios. Existing methods typically treat all OOD samples as a single class, despite real-world applications, especially high-stake settings such as fraud detection and medical diagnosis, demanding deeper insights into OOD samples, including their probable labels. This raises a critical question: can OOD detection be extended to OOD classification without true label information? To address this question, we propose a Coarse-to-Fine open-set Classification (CFC) framework that leverages large language models (LLMs) for graph datasets. CFC consists of three key components: a coarse classifier that uses LLM prompts for OOD detection and outlier label generation, a GNN-based fine classifier trained with OOD samples identified by the coarse classifier for enhanced OOD detection and ID classification, and refined OOD classification achieved through LLM prompts and post-processed OOD labels. Unlike methods that rely on synthetic or auxiliary OOD samples, CFC employs semantic OOD instances that are genuinely out-of-distribution based on their inherent meaning, improving interpretability and practical utility. Experimental results show that CFC improves OOD detection by ten percent over state-of-the-art methods on graph and text domains and achieves up to seventy percent accuracy in OOD classification on graph datasets. --- ### Semi-Supervised Multi-View Crowd Counting by Ranking Multi-View Fusion Models Authors: Qi Zhang, Yunfei Gong, Zhidan Xie, Zhizi Wang, Antoni B. Chan, Hui Huang Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16243v1 Multi-view crowd counting has been proposed to deal with the severe occlusion issue of crowd counting in large and wide scenes. However, due to the difficulty of collecting and annotating multi-view images, the datasets for multi-view counting have a limited number of multi-view frames and scenes. To solve the problem of limited data, one approach is to collect synthetic data to bypass the annotating step, while another is to propose semi- or weakly-supervised or unsupervised methods that demand less multi-view data. In this paper, we propose two semi-supervised multi-view crowd counting frameworks by ranking the multi-view fusion models of different numbers of input views, in terms of the model predictions or the model uncertainties. Specifically, for the first method (vanilla model), we rank the multi-view fusion models' prediction results of different numbers of camera-view inputs, namely, the model's predictions with fewer camera views shall not be larger than the predictions with more camera views. For the second method, we rank the estimated model uncertainties of the multi-view fusion models with a variable number of view inputs, guided by the multi-view fusion models' prediction errors, namely, the model uncertainties with more camera views shall not be larger than those with fewer camera views. These constraints are introduced into the model training in a semi-supervised fashion for multi-view counting with limited labeled data. The experiments demonstrate the advantages of the proposed multi-view model ranking methods compared with other semi-supervised counting methods. --- ### Bayesian Empirical Bayes: Simultaneous Inference from Probabilistic Symmetries Authors: Bohan Wu, Eli N. Weinstein, David M. Blei Published: 2025-12-18 Categories: stat.ME, stat.ML arXiv: https://arxiv.org/abs/2512.16239v3 Empirical Bayes (EB) improves the accuracy of simultaneous inference "by learning from the experience of others" (Efron, 2012). Classical EB theory focuses on latent variables that are iid draws from a fitted prior (Efron, 2019). Modern applications, however, feature complex structure, like arrays, spatial processes, or covariates. We propose a generalized approach to empirical Bayes based on probabilistic symmetry. Our method pairs a simultaneous inference problem with an unknown prior to a symmetry assumption on the joint distribution of the latent variables. Each symmetry implies an ergodic decomposition, which we use to derive a corresponding empirical Bayes method. We call this method Bayesian empirical Bayes (BEB). BEB recovers classical empirical Bayes methods, which implicitly assume exchangeability. We extend EB to other probabilistic symmetries: (i) EB matrix recovery for arrays and graphs; (ii) covariate-informed EB for conditional data; and (iii) EB spatial regression under shift invariance. We develop scalable algorithms based on variational inference and neural networks. In simulations, BEB outperforms existing denoising methods. On real data, we demonstrate BEB on cancer gene-expression and brain-connectivity matrices and NYC air-quality data. --- ### Scaling Spatial Reasoning in MLLMs through Programmatic Data Synthesis Authors: Zhi Helu, Huang Jingjing, Xu Wang, Xu Yangbin, Zhang Wanyue, Jiang Baoyang, Deng Shirui, Zhu Liang, Li Fangfang, Zhao Tiejun, Lin Yankai, Yao Yuan Published: 2025-12-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.16237v1 Embodied intelligence, a grand challenge in artificial intelligence, is fundamentally constrained by the limited spatial understanding and reasoning capabilities of current models. Prevailing efforts to address this through enhancing Vision-Language Models (VLMs) are trapped in a dilemma: template-based datasets are scalable but structurally rigid, while manual annotation is linguistically diverse but unscalable and, critically, computationally imprecise. We introduce SPRITE, a novel framework that overcomes this dilemma by leveraging simulators and large models to programmatically synthesize scalable, diverse, and high-quality spatial reasoning data. The core innovation of SPRITE is to reframe ground-truth generation as a code-generation task. We utilize LLMs to compile complex spatial questions into executable programs, which are then verified against high-precision scene meta-information extracted from simulators. This ensures our ground truth is both computationally precise and verifiable, while the generative power of LLMs provides vast linguistic diversity. Leveraging this pipeline, we have curated a dataset encompassing 3 simulators, 11k+ scenes, and 300k+ image/video instruction-tuning pairs. We demonstrate that a VLM trained on our data achieves significant performance gains on multiple spatial benchmarks and outperforms other open-source datasets of equivalent size. Furthermore, a scalability analysis confirms our hypothesis that overcoming the low-diversity nature of traditional template methods is essential for building robust, generalizable spatial intelligence. We will make the SPRITE framework code and the full 300k+ dataset publicly available to facilitate future research in spatial intelligence. --- ### The Evolution of Reranking Models in Information Retrieval: From Heuristic Methods to Large Language Models Authors: Tejul Pandit, Sakshi Mahendru, Meet Raval, Dhvani Upadhyay Published: 2025-12-18 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2512.16236v1 Reranking is a critical stage in contemporary information retrieval (IR) systems, improving the relevance of the user-presented final results by honing initial candidate sets. This paper is a thorough guide to examine the changing reranker landscape and offer a clear view of the advancements made in reranking methods. We present a comprehensive survey of reranking models employed in IR, particularly within modern Retrieval Augmented Generation (RAG) pipelines, where retrieved documents notably influence output quality. We embark on a chronological journey through the historical trajectory of reranking techniques, starting with foundational approaches, before exploring the wide range of sophisticated neural network architectures such as cross-encoders, sequence-generation models like T5, and Graph Neural Networks (GNNs) utilized for structural information. Recognizing the computational cost of advancing neural rerankers, we analyze techniques for enhancing efficiency, notably knowledge distillation for creating competitive, lighter alternatives. Furthermore, we map the emerging territory of integrating Large Language Models (LLMs) in reranking, examining novel prompting strategies and fine-tuning tactics. This survey seeks to elucidate the fundamental ideas, relative effectiveness, computational features, and real-world trade-offs of various reranking strategies. The survey provides a structured synthesis of the diverse reranking paradigms, highlighting their underlying principles and comparative strengths and weaknesses. --- ### AI-Powered Dermatological Diagnosis: From Interpretable Models to Clinical Implementation A Comprehensive Framework for Accessible and Trustworthy Skin Disease Detection Authors: Satya Narayana Panda, Vaishnavi Kukkala, Spandana Iyer Published: 2025-12-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.16235v1 Dermatological conditions affect 1.9 billion people globally, yet accurate diagnosis remains challenging due to limited specialist availability and complex clinical presentations. Family history significantly influences skin disease susceptibility and treatment responses, but is often underutilized in diagnostic processes. This research addresses the critical question: How can AI-powered systems integrate family history data with clinical imaging to enhance dermatological diagnosis while supporting clinical trial validation and real-world implementation? We developed a comprehensive multi-modal AI framework that combines deep learning-based image analysis with structured clinical data, including detailed family history patterns. Our approach employs interpretable convolutional neural networks integrated with clinical decision trees that incorporate hereditary risk factors. The methodology includes prospective clinical trials across diverse healthcare settings to validate AI-assisted diagnosis against traditional clinical assessment. In this work, validation was conducted with healthcare professionals to assess AI-assisted outputs against clinical expectations; prospective clinical trials across diverse healthcare settings are proposed as future work. The integrated AI system demonstrates enhanced diagnostic accuracy when family history data is incorporated, particularly for hereditary skin conditions such as melanoma, psoriasis, and atopic dermatitis. Expert feedback indicates potential for improved early detection and more personalized recommendations; formal clinical trials are planned. The framework is designed for integration into clinical workflows while maintaining interpretability through explainable AI mechanisms. --- ### ARMFlow: AutoRegressive MeanFlow for Online 3D Human Reaction Generation Authors: Zichen Geng, Zeeshan Hayder, Wei Liu, Hesheng Wang, Ajmal Mian Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16234v2 3D human reaction generation faces three main challenges:(1) high motion fidelity, (2) real-time inference, and (3) autoregressive adaptability for online scenarios. Existing methods fail to meet all three simultaneously. We propose ARMFlow, a MeanFlow-based autoregressive framework that models temporal dependencies between actor and reactor motions. It consists of a causal context encoder and an MLP-based velocity predictor. We introduce Bootstrap Contextual Encoding (BSCE) in training, encoding generated history instead of the ground-truth ones, to alleviate error accumulation in autoregressive generation. We further introduce the offline variant ReMFlow, achieving state-of-the-art performance with the fastest inference among offline methods. Our ARMFlow addresses key limitations of online settings by: (1) enhancing semantic alignment via a global contextual encoder; (2) achieving high accuracy and low latency in a single-step inference; and (3) reducing accumulated errors through BSCE. Our single-step online generation surpasses existing online methods on InterHuman and InterX by about 30% in FID, while matching offline state-of-the-art performance despite using only partial sequence conditions. --- ### DAG Learning from Zero-Inflated Count Data Using Continuous Optimization Authors: Noriaki Sato, Marco Scutari, Shuichi Kawano, Rui Yamaguchi, Seiya Imoto Published: 2025-12-18 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2512.16233v1 We address network structure learning from zero-inflated count data by casting each node as a zero-inflated generalized linear model and optimizing a smooth, score-based objective under a directed acyclic graph constraint. Our Zero-Inflated Continuous Optimization (ZICO) approach uses node-wise likelihoods with canonical links and enforces acyclicity through a differentiable surrogate constraint combined with sparsity regularization. ZICO achieves superior performance with faster runtimes on simulated data. It also performs comparably to or better than common algorithms for reverse engineering gene regulatory networks. ZICO is fully vectorized and mini-batched, enabling learning on larger variable sets with practical runtimes in a wide range of domains. --- ### LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding Authors: Chenkai Xu, Yijie Jin, Jiajun Li, Yi Tu, Guoping Long, Dandan Tu, Mingcong Song, Hongjie Si, Tianqi Hou, Junchi Yan, Zhijie Deng Published: 2025-12-18 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.16229v2 Diffusion Large Language Models (dLLMs) have demonstrated significant potential for high-speed inference. However, current confidence-driven decoding strategies are constrained by limited parallelism, typically achieving only 1--3 tokens per forward pass (TPF). In this work, we identify that the degree of parallelism during dLLM inference is highly sensitive to the Token Filling Order (TFO). Then, we introduce Lookahead PArallel Decoding LoPA, a training-free, plug-and-play algorithm, to identify a superior TFO and hence accelerate inference. LoPA concurrently explores distinct candidate TFOs via parallel branches, and selects the one with the highest potential for future parallelism based on branch confidence. We apply LoPA to the state-of-the-art D2F model and observe a substantial enhancement in decoding efficiency. Notably, LoPA increases the TPF of D2F-Dream to 10.1 on the GSM8K while maintaining performance superior to the Dream baseline. Furthermore, to facilitate this unprecedented degree of parallelism, we develop a specialized multi-device inference system featuring Branch Parallelism (BP), which achieves a single-sample throughput of 1073.9 tokens per second under multi-GPU deployment. The code is available at https://github.com/zhijie-group/LoPA. --- ### An Information-Theoretic Framework for Robust Large Language Model Editing Authors: Qizhou Chen, Chengyu Wang, Taolin Zhang, Xiaofeng He Published: 2025-12-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2512.16227v1 Large Language Models (LLMs) have become indispensable tools in science, technology, and society, enabling transformative advances across diverse fields. However, errors or outdated information within these models can undermine their accuracy and restrict their safe deployment. Developing efficient strategies for updating model knowledge without the expense and disruption of full retraining remains a critical challenge. Current model editing techniques frequently struggle to generalize corrections beyond narrow domains, leading to unintended consequences and limiting their practical impact. Here, we introduce a novel framework for editing LLMs, grounded in information bottleneck theory. This approach precisely compresses and isolates the essential information required for generalizable knowledge correction while minimizing disruption to unrelated model behaviors. Building upon this foundation, we present the Information Bottleneck Knowledge Editor (IBKE), which leverages compact latent representations to guide gradient-based updates, enabling robust and broadly applicable model editing. We validate IBKE's effectiveness across multiple LLM architectures and standard benchmark tasks, demonstrating state-of-the-art accuracy and improved generality and specificity of edits. These findings establish a theoretically principled and practical paradigm for open-domain knowledge editing, advancing the utility and trustworthiness of LLMs in real-world applications. --- ### Image Compression Using Singular Value Decomposition Authors: Justin Jiang Published: 2025-12-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.16226v1 Images are a substantial portion of the internet, making efficient compression important for reducing storage and bandwidth demands. This study investigates the use of Singular Value Decomposition and low-rank matrix approximations for image compression, evaluating performance using relative Frobenius error and compression ratio. The approach is applied to both grayscale and multichannel images to assess its generality. Results show that the low-rank approximations often produce images that appear visually similar to the originals, but the compression efficiency remains consistently worse than established formats such as JPEG, JPEG2000, and WEBP at comparable error levels. At low tolerated error levels, the compressed representation produced by Singular Value Decomposition can even exceed the size of the original image, indicating that this method is not competitive with industry-standard codecs for practical image compression. --- ### Neural emulation of gravity-driven geohazard runout Authors: Lorenzo Nava, Ye Chen, Maximillian Van Wyk de Vries Published: 2025-12-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2512.16221v1 Predicting geohazard runout is critical for protecting lives, infrastructure and ecosystems. Rapid mass flows, including landslides and avalanches, cause several thousand deaths across a wide range of environments, often travelling many kilometres from their source. The wide range of source conditions and material properties governing these flows makes their runout difficult to anticipate, particularly for downstream communities that may be suddenly exposed to severe impacts. Accurately predicting runout at scale requires models that are both physically realistic and computationally efficient, yet existing approaches face a fundamental speed-realism trade-off. Here we train a machine learning model to predict geohazard runout across representative real world terrains. The model predicts both flow extent and deposit thickness with high accuracy and 100 to 10,000 times faster computation than numerical solvers. It is trained on over 100,000 numerical simulations across over 10,000 real world digital elevation model chips and reproduces key physical behaviours, including avulsion and deposition patterns, while generalizing across different flow types, sizes and landscapes. Our results demonstrate that neural emulation enables rapid, spatially resolved runout prediction across diverse real world terrains, opening new opportunities for disaster risk reduction and impact-based forecasting. These results highlight neural emulation as a promising pathway for extending physically realistic geohazard modelling to spatial and temporal scales relevant for large scale early warning systems. ---