# Research Papers — 2025-12 · bundle page 95 of 175 > Full-text markdown of papers #4701–4750 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/95 [← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/94.md) [Next bundle →](https://aipod.dev/papers/archive/2025-12/page/96.md) ### Persistent Backdoor Attacks under Continual Fine-Tuning of LLMs Authors: Jing Cui, Yufei Han, Jianbin Jiao, Junge Zhang Published: 2025-12-12 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2512.14741v1 Backdoor attacks embed malicious behaviors into Large Language Models (LLMs), enabling adversaries to trigger harmful outputs or bypass safety controls. However, the persistence of the implanted backdoors under user-driven post-deployment continual fine-tuning has been rarely examined. Most prior works evaluate the effectiveness and generalization of implanted backdoors only at releasing and empirical evidence shows that naively injected backdoor persistence degrades after updates. In this work, we study whether and how implanted backdoors persist through a multi-stage post-deployment fine-tuning. We propose P-Trojan, a trigger-based attack algorithm that explicitly optimizes for backdoor persistence across repeated updates. By aligning poisoned gradients with those of clean tasks on token embeddings, the implanted backdoor mapping is less likely to be suppressed or forgotten during subsequent updates. Theoretical analysis shows the feasibility of such persistent backdoor attacks after continual fine-tuning. And experiments conducted on the Qwen2.5 and LLaMA3 families of LLMs, as well as diverse task sequences, demonstrate that P-Trojan achieves over 99% persistence while preserving clean-task accuracy. Our findings highlight the need for persistence-aware evaluation and stronger defenses in realistic model adaptation pipelines. --- ### NoveltyRank: A Retrieval-Augmented Framework for Conceptual Novelty Estimation in AI Research Authors: Zhengxu Yan, Han Li, Yuming Feng Published: 2025-12-12 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2512.14738v2 The accelerating pace of scientific publication makes it difficult to identify truly original research among incremental work. We propose a framework for estimating the conceptual novelty of research papers by combining semantic representation learning with retrieval-based comparison against prior literature. We model novelty as both a binary classification task (novel vs. non-novel) and a pairwise ranking task (comparative novelty), enabling absolute and relative assessments. Experiments benchmark three model scales, ranging from compact domain-specific encoders to a zero-shot frontier model. Results show that fine-tuned lightweight models outperform larger zero-shot models despite their smaller parameter count, indicating that task-specific supervision matters more than scale for conceptual novelty estimation. We further deploy the best-performing model as an online system for public interaction and real-time novelty scoring. --- ### GraphPerf-RT: A Graph-Driven Performance Model for Hardware-Aware Scheduling of OpenMP Codes Authors: Mohammad Pivezhandi, Mahdi Banisharif, Saeed Bakhshan, Abusayeed Saifullah, Ali Jannesari Published: 2025-12-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.12091v3 Autonomous AI agents on embedded platforms require real-time, risk-aware scheduling under resource and thermal constraints. Classical heuristics struggle with workload irregularity, tabular regressors discard structural information, and model-free reinforcement learning (RL) risks overheating. We introduce GraphPerf-RT, a graph neural network surrogate achieving deep learning accuracy at heuristic speeds (2-7ms). GraphPerf-RT is, to our knowledge, the first to unify task DAG topology, CFG-derived code semantics, and runtime context (per-core DVFS, thermal state, utilization) in a heterogeneous graph with typed edges encoding precedence, placement, and contention. Evidential regression with Normal-Inverse-Gamma priors provides calibrated uncertainty; we validate on makespan prediction for risk-aware scheduling. Experiments on three ARM platforms (Jetson TX2, Orin NX, RUBIK Pi) achieve R^2 = 0.81 on log-transformed makespan with Spearman rho = 0.95 and conservative uncertainty calibration (PICP = 99.9% at 95% confidence). Integration with four RL methods demonstrates that multi-agent model-based RL with GraphPerf-RT as the world model achieves 66% makespan reduction and 82% energy reduction versus model-free baselines, with zero thermal violations. --- ### SPDMark: Selective Parameter Displacement for Robust Video Watermarking Authors: Samar Fares, Nurbek Tastan, Karthik Nandakumar Published: 2025-12-12 Categories: cs.CV, cs.CR, cs.LG arXiv: https://arxiv.org/abs/2512.12090v2 The advent of high-quality video generation models has amplified the need for robust watermarking schemes that can be used to reliably detect and track the provenance of generated videos. Existing video watermarking methods based on both post-hoc and in-generation approaches fail to simultaneously achieve imperceptibility, robustness, and computational efficiency. This work introduces a novel framework for in-generation video watermarking called SPDMark (pronounced `SpeedMark') based on selective parameter displacement of a video diffusion model. Watermarks are embedded into the generated videos by modifying a subset of parameters in the generative model. To make the problem tractable, the displacement is modeled as an additive composition of layer-wise basis shifts, where the final composition is indexed by the watermarking key. For parameter efficiency, this work specifically leverages low-rank adaptation (LoRA) to implement the basis shifts. During the training phase, the basis shifts and the watermark extractor are jointly learned by minimizing a combination of message recovery, perceptual similarity, and temporal consistency losses. To detect and localize temporal modifications in the watermarked videos, we use a cryptographic hashing function to derive frame-specific watermark messages from the given base watermarking key. During watermark extraction, maximum bipartite matching is applied to recover the correct frame order, even from temporally tampered videos. Evaluations on both text-to-video and image-to-video generation models demonstrate the ability of SPDMark to generate imperceptible watermarks that can be recovered with high accuracy and also establish its robustness against a variety of common video modifications. --- ### VEGAS: Mitigating Hallucinations in Large Vision-Language Models via Vision-Encoder Attention Guided Adaptive Steering Authors: Zihu Wang, Boxun Xu, Yuxuan Xia, Peng Li Published: 2025-12-12 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2512.12089v1 Large vision-language models (LVLMs) exhibit impressive ability to jointly reason over visual and textual inputs. However, they often produce outputs that are linguistically fluent but factually inconsistent with the visual evidence, i.e., they hallucinate. Despite growing efforts to mitigate such hallucinations, a key question remains: what form of visual attention can effectively suppress hallucinations during decoding? In this work, we provide a simple answer: the vision encoder's own attention map. We show that LVLMs tend to hallucinate when their final visual-attention maps fail to concentrate on key image objects, whereas the vision encoder's more concentrated attention maps substantially reduce hallucinations. To further investigate the cause, we analyze vision-text conflicts during decoding and find that these conflicts peak in the language model's middle layers. Injecting the vision encoder's attention maps into these layers effectively suppresses hallucinations. Building on these insights, we introduce VEGAS, a simple yet effective inference-time method that integrates the vision encoder's attention maps into the language model's mid-layers and adaptively steers tokens which fail to concentrate on key image objects. Extensive experiments across multiple benchmarks demonstrate that VEGAS consistently achieves state-of-the-art performance in reducing hallucinations. --- ### Reliable Policy Iteration: Performance Robustness Across Architecture and Environment Perturbations Authors: S. R. Eshwar, Aniruddha Mukherjee, Kintan Saha, Krishna Agarwal, Gugan Thoppe, Aditya Gopalan, Gal Dalal Published: 2025-12-12 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.12088v1 In a recent work, we proposed Reliable Policy Iteration (RPI), that restores policy iteration's monotonicity-of-value-estimates property to the function approximation setting. Here, we assess the robustness of RPI's empirical performance on two classical control tasks -- CartPole and Inverted Pendulum -- under changes to neural network and environmental parameters. Relative to DQN, Double DQN, DDPG, TD3, and PPO, RPI reaches near-optimal performance early and sustains this policy as training proceeds. Because deep RL methods are often hampered by sample inefficiency, training instability, and hyperparameter sensitivity, our results highlight RPI's promise as a more reliable alternative. --- ### BLASST: Dynamic BLocked Attention Sparsity via Softmax Thresholding Authors: Jiayi Yuan, Cameron Shinn, Kai Xu, Jingze Cui, George Klimiashvili, Guangxuan Xiao, Perkz Zheng, Bo Li, Yuxin Zhou, Zhouhai Ye, Weijie You, Tian Zheng et al. Published: 2025-12-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.12087v3 The growing demand for long-context inference capabilities in Large Language Models (LLMs) has intensified the computational and memory bottlenecks inherent to the self-attention mechanism. To address this challenge, we introduce BLASST, a drop-in, dynamic sparse attention mechanism that accelerates inference by using only a fixed scalar threshold to skip attention blocks. Our method targets practical inference deployment by removing the barriers to adoption present in existing works. As such, BLASST eliminates training requirements, avoids expensive pre-computation passes, accelerates both prefill and decode across all major attention variants (MHA, GQA, MQA, and MLA), provides optimized support for modern hardware, and easily integrates into existing frameworks. This is achieved by reusing online softmax statistics to identify negligible attention scores, skipping softmax, value block loads, and the subsequent matrix multiplication. We demonstrate the BLASST algorithm by delivering optimized kernels with negligible latency overhead. Our automated threshold calibration procedure reveals a simple inverse relationship between optimal threshold and context length, meaning we require only a single threshold each for prefill and decode per model. Preserving benchmark accuracy, we demonstrate a 1.52x speedup for prefill at 71.9% sparsity and a 1.48x speedup for decode at 73.2% sparsity on modern GPUs. --- ### CLOAK: Contrastive Guidance for Latent Diffusion-Based Data Obfuscation Authors: Xin Yang, Omid Ardakanian Published: 2025-12-12 Categories: cs.LG, cs.CR arXiv: https://arxiv.org/abs/2512.12086v2 Data obfuscation is a promising technique for mitigating attribute inference attacks by semi-trusted parties with access to time-series data emitted by sensors. Recent advances leverage conditional generative models together with adversarial training or mutual information-based regularization to balance data privacy and utility. However, these methods often require modifying the downstream task, struggle to achieve a satisfactory privacy-utility trade-off, or are computationally intensive, making them impractical for deployment on resource-constrained mobile IoT devices. We propose Cloak, a novel data obfuscation framework based on latent diffusion models. In contrast to prior work, we employ contrastive learning to extract disentangled representations, which guide the latent diffusion process to retain useful information while concealing private information. This approach enables users with diverse privacy needs to navigate the privacy-utility trade-off with minimal retraining. Extensive experiments on four public time-series datasets, spanning multiple sensing modalities, and a dataset of facial images demonstrate that Cloak consistently outperforms state-of-the-art obfuscation techniques, reducing utility loss by up to 7.21% and privacy loss by up to 5.76%, and is well-suited for deployment in resource-constrained settings. --- ### RePack then Refine: Efficient Diffusion Transformer with Vision Foundation Model Authors: Guanfang Dong, Luke Schultz, Negar Hassanpour, Chao Gao Published: 2025-12-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.12083v3 Semantic-rich features from Vision Foundation Models (VFMs) have been leveraged to enhance Latent Diffusion Models (LDMs). However, raw VFM features are typically high-dimensional and redundant, increasing the difficulty of learning and reducing training efficiency for Diffusion Transformers (DiTs). In this paper, we propose Repack then Refine, a three-stage framework that brings the semantic-rich VFM features to DiT while further accelerating learning efficiency. Specifically, the RePack module projects the high-dimensional features onto a compact, low-dimensional manifold. This filters out the redundancy while preserving essential structural information. A standard DiT is then trained for generative modeling on this highly compressed latent space. Finally, to restore the high-frequency details lost due to the compression in RePack, we propose a Latent-Guided Refiner, which is trained lastly for enhancing the image details. On ImageNet-1K, RePack-DiT-XL/1 achieves an FID of 1.82 in only 64 training epochs. With the Refiner module, performance further improves to an FID of 1.65, significantly surpassing latest LDMs in terms of convergence efficiency. Our results demonstrate that packing VFM features, followed by targeted refinement, is a highly effective strategy for balancing generative fidelity with training efficiency. Source code is publicly available at https://github.com/guanfangdong/RePack-then-Refine. --- ### Congestion Reduction in EV Charger Placement Using Traffic Equilibrium Models Authors: Semih Kara, Yasin Sonmez, Can Kizilkale, Alex Kurzhanskiy, Nuno C. Martins, Murat Arcak Published: 2025-12-12 Categories: eess.SY, cs.AI, cs.SI, math.OC arXiv: https://arxiv.org/abs/2512.12081v1 Growing EV adoption can worsen traffic conditions if chargers are sited without regard to their impact on congestion. We study how to strategically place EV chargers to reduce congestion using two equilibrium models: one based on congestion games and one based on an atomic queueing simulation. We apply both models within a scalable greedy station-placement algorithm. Experiments show that this greedy scheme yields optimal or near-optimal congestion outcomes in realistic networks, even though global optimality is not guaranteed as we show with a counterexample. We also show that the queueing-based approach yields more realistic results than the congestion-game model, and we present a unified methodology that calibrates congestion delays from queue simulation and solves equilibrium in link-space. --- ### BAgger: Backwards Aggregation for Mitigating Drift in Autoregressive Video Diffusion Models Authors: Ryan Po, Eric Ryan Chan, Changan Chen, Gordon Wetzstein Published: 2025-12-12 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2512.12080v1 Autoregressive video models are promising for world modeling via next-frame prediction, but they suffer from exposure bias: a mismatch between training on clean contexts and inference on self-generated frames, causing errors to compound and quality to drift over time. We introduce Backwards Aggregation (BAgger), a self-supervised scheme that constructs corrective trajectories from the model's own rollouts, teaching it to recover from its mistakes. Unlike prior approaches that rely on few-step distillation and distribution-matching losses, which can hurt quality and diversity, BAgger trains with standard score or flow matching objectives, avoiding large teachers and long-chain backpropagation through time. We instantiate BAgger on causal diffusion transformers and evaluate on text-to-video, video extension, and multi-prompt generation, observing more stable long-horizon motion and better visual consistency with reduced drift. --- ### SigTime: Learning and Visually Explaining Time Series Signatures Authors: Yu-Chia Huang, Juntong Chen, Dongyu Liu, Kwan-Liu Ma Published: 2025-12-12 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2512.12076v1 Understanding and distinguishing temporal patterns in time series data is essential for scientific discovery and decision-making. For example, in biomedical research, uncovering meaningful patterns in physiological signals can improve diagnosis, risk assessment, and patient outcomes. However, existing methods for time series pattern discovery face major challenges, including high computational complexity, limited interpretability, and difficulty in capturing meaningful temporal structures. To address these gaps, we introduce a novel learning framework that jointly trains two Transformer models using complementary time series representations: shapelet-based representations to capture localized temporal structures and traditional feature engineering to encode statistical properties. The learned shapelets serve as interpretable signatures that differentiate time series across classification labels. Additionally, we develop a visual analytics system -- SigTIme -- with coordinated views to facilitate exploration of time series signatures from multiple perspectives, aiding in useful insights generation. We quantitatively evaluate our learning framework on eight publicly available datasets and one proprietary clinical dataset. Additionally, we demonstrate the effectiveness of our system through two usage scenarios along with the domain experts: one involving public ECG data and the other focused on preterm labor analysis. --- ### Evaluation of Sampling Strategies and Physics-Informed Kolmogorov--Arnold Networks in Unbounded Domains Authors: Gregorio Pérez-Bernal, Oscar Rincón-Cardeño, Silvana Montoya-Noguera, Nicolás Guarín-Zapata Published: 2025-12-12 Categories: cs.LG, math-ph arXiv: https://arxiv.org/abs/2512.12074v2 Physics-informed neural networks (PINNs) have emerged as an effective approach for solving partial differential equations (PDEs) by incorporating physical laws into the learning process. However, their application to infinite and semi-infinite domains remains challenging due to the difficulty of representing unbounded regions with a finite number of training points. This work evaluates physics-informed learning strategies for inverse PDE problems in unbounded domains, focusing on the influence of sampling strategies based on uniform, Gaussian, and exponential distributions, and on the use of conventional multilayer perceptrons (MLPs) and Kolmogorov--Arnold Networks (KANs) within the PINN framework. The proposed benchmark considers manufactured inverse problems on infinite and semi-infinite domains, enabling quantitative assessment of reconstruction accuracy and computational efficiency. Results show that Gaussian and exponential sampling, controlled through explicit distribution parameters, improve reconstruction accuracy in the far field and offer a direct mechanism for embedding problem-specific prior knowledge into training. PIKANs, whose edge-based functional representation is grounded in the Kolmogorov--Arnold representation theorem, match or exceed the accuracy of MLP-based PINNs while requiring fewer neurons, though their spline-based formulation increases the computational cost of training, inference, and derivative evaluation. --- ### VOYAGER: A Training Free Approach for Generating Diverse Datasets using LLMs Authors: Avinash Amballa, Yashas Malur Saidutta, Chi-Heng Lin, Vivek Kulkarni, Srinivas Chappidi Published: 2025-12-12 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2512.12072v2 Large language models (LLMs) are increasingly being used to generate synthetic datasets for the evaluation and training of downstream models. However, prior work has noted that such generated data lacks diversity. In this paper, we propose Voyager, a novel principled approach to generate diverse datasets. Our approach is iterative and directly optimizes a mathematical quantity that optimizes the diversity of the dataset using the machinery of determinantal point processes. Furthermore, our approach is training-free, applicable to closed-source models, and scalable. In addition to providing theoretical justification for the working of our method, we also demonstrate through comprehensive experiments that Voyager significantly outperforms popular baseline approaches by providing a 1.5-3 times improvement in diversity. --- ### Rethinking Jailbreak Detection of Large Vision Language Models with Representational Contrastive Scoring Authors: Peichun Hua, Hao Li, Shanghao Shi, Zhiyuan Yu, Ning Zhang Published: 2025-12-12 Categories: cs.CR, cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2512.12069v3 Large Vision-Language Models (LVLMs) are vulnerable to a growing array of multimodal jailbreak attacks, necessitating defenses that are both generalizable to novel threats and efficient for practical deployment. Many current strategies fall short, either targeting specific attack patterns, which limits generalization, or imposing high computational overhead. While lightweight anomaly-detection methods offer a promising direction, we find that their common one-class design tends to confuse unseen benign inputs with malicious ones, leading to unreliable over-rejection. To address this, we propose Representational Contrastive Scoring (RCS), a framework built on a key insight: the most potent safety signals reside within the LVLM's own internal representations. Our approach inspects the internal geometry of these representations, learning a lightweight projection to maximally separate benign and malicious inputs in safety-critical layers. This enables a simple yet powerful contrastive score that differentiates true malicious intent from mere distribution shift. Our instantiations, MCD (Mahalanobis Contrastive Detection) and KCD (K-nearest Contrastive Detection), achieve state-of-the-art performance on a challenging evaluation protocol designed to test generalization to unseen attack types. This work demonstrates that effective jailbreak detection can be achieved by applying simple, interpretable statistical methods to the internal representations, offering a practical path towards safer LVLM deployment. --- ### The Instability of Safety: How Random Seeds and Temperature Expose Inconsistent LLM Refusal Behavior Authors: Erik Larsen Published: 2025-12-12 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2512.12066v4 Current safety evaluations of large language models rely on single-shot testing, implicitly assuming that model responses are deterministic and representative of the model's safety alignment. We challenge this assumption by investigating the stability of safety refusal decisions across random seeds and temperature settings. Testing four instruction-tuned models from three families (Llama 3.1 8B, Qwen 2.5 7B, Qwen 3 8B, Gemma 3 12B) on 876 harmful prompts across 20 sampling configurations (4 temperatures x 5 seeds), we find that 18-28% of prompts exhibit decision flips (the REFUSE/PARTIAL/COMPLY label changes across configurations), depending on the model. Our Safety Stability Index (SSI) reveals that higher temperatures significantly reduce decision stability (Friedman chi-squared = 396.81, p < 0.001), with mean within-temperature SSI dropping from 0.977 at temperature 0.0 to 0.942 at temperature 1.0. We validate findings across all model families using Claude 3.5 Haiku as a unified external judge, achieving 89.1% inter-judge agreement with the Llama 70B judge on the two models both judges labeled (Cohen's kappa = 0.62). Within each model, prompts with higher compliance rates exhibit lower stability (Spearman rho = -0.47 to -0.70, all p < 0.001), indicating that models "waver" more on borderline requests. These findings demonstrate that single-shot safety evaluations are insufficient for reliable safety assessment and that evaluation protocols must account for stochastic variation in model behavior. For Llama 3.1 8B, single-shot evaluation agrees with multi-sample ground truth only 92.5% of the time when pooling across temperatures (98.7% at greedy to 90.3% at temperature 1.0), and we recommend scaling samples with temperature--one at greedy, three at low temperature, more at higher temperatures (where three reach only ~95%), and ten when pooling--rather than a single flat threshold. --- ### Instruction-Tuning Open-Weight Language Models for BPMN Model Generation Authors: Gökberk Çelikmasat, Atay Özgövde, Fatma Başak Aydemir Published: 2025-12-12 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2512.12063v1 Domain models are central to software engineering, as they enable a shared understanding, guide implementation, and support automated analyses and model-driven development. Yet, despite these benefits, practitioners often skip modeling because it is time-consuming and demands scarce expertise. We address this barrier by investigating whether open-weight large language models, adapted via instruction tuning, can generate high-quality BPMN process models directly from natural language descriptions in a cost-effective and privacy-preserving way. We introduce InstruBPM, a reproducible approach that prepares paired text-diagram data and instruction tunes an open source large language model with parameter-efficient fine-tuning and quantization for on-prem deployment. We evaluate the tuned model through complementary perspectives: (i) text/code similarity using BLEU, ROUGE-L, and METEOR, (ii) structural fidelity using Relative Graph Edit Distance, (iii) guidelines conformance using external tool checks, and (iv) a small expert review. Using a curated subset of a multi-domain BPMN dataset, we compare the tuned model with untuned open-weight baselines and strong proprietary models under consistent prompting regimes. Our compact tuned model outperforms all baselines across sequence and structural metrics while requiring substantially fewer resources; guideline analysis and expert feedback further indicate that the generated diagrams largely follow BPMN best practices and are useful starting points that reduce modeling effort. Overall, instruction tuning improves structural accuracy and robustness compared to untuned baselines and reduces reliance on heavy prompt scaffolding. We publicly share the trained models and scripts to support reproducibility and further research. --- ### CreativeVR: Diffusion-Prior-Guided Approach for Structure and Motion Restoration in Generative and Real Videos Authors: Tejas Panambur, Ishan Rajendrakumar Dave, Chongjian Ge, Ersin Yumer, Xue Bai Published: 2025-12-12 Categories: cs.CV, cs.LG, cs.MM arXiv: https://arxiv.org/abs/2512.12060v1 Modern text-to-video (T2V) diffusion models can synthesize visually compelling clips, yet they remain brittle at fine-scale structure: even state-of-the-art generators often produce distorted faces and hands, warped backgrounds, and temporally inconsistent motion. Such severe structural artifacts also appear in very low-quality real-world videos. Classical video restoration and super-resolution (VR/VSR) methods, in contrast, are tuned for synthetic degradations such as blur and downsampling and tend to stabilize these artifacts rather than repair them, while diffusion-prior restorers are usually trained on photometric noise and offer little control over the trade-off between perceptual quality and fidelity. We introduce CreativeVR, a diffusion-prior-guided video restoration framework for AI-generated (AIGC) and real videos with severe structural and temporal artifacts. Our deep-adapter-based method exposes a single precision knob that controls how strongly the model follows the input, smoothly trading off between precise restoration on standard degradations and stronger structure- and motion-corrective behavior on challenging content. Our key novelty is a temporally coherent degradation module used during training, which applies carefully designed transformations that produce realistic structural failures. To evaluate AIGC-artifact restoration, we propose the AIGC54 benchmark with FIQA, semantic and perceptual metrics, and multi-aspect scoring. CreativeVR achieves state-of-the-art results on videos with severe artifacts and performs competitively on standard video restoration benchmarks, while running at practical throughput (about 13 FPS at 720p on a single 80-GB A100). Project page: https://daveishan.github.io/creativevr-webpage/. --- ### The Forecast Critic: Leveraging Large Language Models for Poor Forecast Identification Authors: Luke Bhan, Hanyu Zhang, Andrew Gordon Wilson, Michael W. Mahoney, Chuck Arvin Published: 2025-12-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.12059v1 Monitoring forecasting systems is critical for customer satisfaction, profitability, and operational efficiency in large-scale retail businesses. We propose The Forecast Critic, a system that leverages Large Language Models (LLMs) for automated forecast monitoring, taking advantage of their broad world knowledge and strong ``reasoning'' capabilities. As a prerequisite for this, we systematically evaluate the ability of LLMs to assess time series forecast quality, focusing on three key questions. (1) Can LLMs be deployed to perform forecast monitoring and identify obviously unreasonable forecasts? (2) Can LLMs effectively incorporate unstructured exogenous features to assess what a reasonable forecast looks like? (3) How does performance vary across model sizes and reasoning capabilities, measured across state-of-the-art LLMs? We present three experiments, including on both synthetic and real-world forecasting data. Our results show that LLMs can reliably detect and critique poor forecasts, such as those plagued by temporal misalignment, trend inconsistencies, and spike errors. The best-performing model we evaluated achieves an F1 score of 0.88, somewhat below human-level performance (F1 score: 0.97). We also demonstrate that multi-modal LLMs can effectively incorporate unstructured contextual signals to refine their assessment of the forecast. Models correctly identify missing or spurious promotional spikes when provided with historical context about past promotions (F1 score: 0.84). Lastly, we demonstrate that these techniques succeed in identifying inaccurate forecasts on the real-world M5 time series dataset, with unreasonable forecasts having an sCRPS at least 10% higher than that of reasonable forecasts. These findings suggest that LLMs, even without domain-specific fine-tuning, may provide a viable and scalable option for automated forecast monitoring and evaluation. --- ### Enhancing deep learning performance on burned area delineation from SPOT-6/7 imagery for emergency management Authors: Maria Rodriguez, Minh-Tan Pham, Martin Sudmanns, Quentin Poterek, Oscar Narvaez Published: 2025-12-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.12056v1 After a wildfire, delineating burned areas (BAs) is crucial for quantifying damages and supporting ecosystem recovery. Current BA mapping approaches rely on computer vision models trained on post-event remote sensing imagery, but often overlook their applicability to time-constrained emergency management scenarios. This study introduces a supervised semantic segmentation workflow aimed at boosting both the performance and efficiency of BA delineation. It targets SPOT-6/7 imagery due to its very high resolution and on-demand availability. Experiments are evaluated based on Dice score, Intersection over Union, and inference time. The results show that U-Net and SegFormer models perform similarly with limited training data. However, SegFormer requires more resources, challenging its practical use in emergencies. Incorporating land cover data as an auxiliary task enhances model robustness without increasing inference time. Lastly, Test-Time Augmentation improves BA delineation performance but raises inference time, which can be mitigated with optimization methods like Mixed Precision. --- ### Adaptive federated learning for ship detection across diverse satellite imagery sources Authors: Tran-Vu La, Minh-Tan Pham, Yu Li, Patrick Matgen, Marco Chini Published: 2025-12-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.12053v1 We investigate the application of Federated Learning (FL) for ship detection across diverse satellite datasets, offering a privacy-preserving solution that eliminates the need for data sharing or centralized collection. This approach is particularly advantageous for handling commercial satellite imagery or sensitive ship annotations. Four FL models including FedAvg, FedProx, FedOpt, and FedMedian, are evaluated and compared to a local training baseline, where the YOLOv8 ship detection model is independently trained on each dataset without sharing learned parameters. The results reveal that FL models substantially improve detection accuracy over training on smaller local datasets and achieve performance levels close to global training that uses all datasets during the training. Furthermore, the study underscores the importance of selecting appropriate FL configurations, such as the number of communication rounds and local training epochs, to optimize detection precision while maintaining computational efficiency. --- ### Context-Aware Agentic Power Resources Optimisation in EV using Smart2ChargeApp Authors: Muddsair Sharif, Huseyin Seker Published: 2025-12-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.12048v1 This paper presents a novel context-sensitive multi\-agent coordination for dynamic resource allocation (CAMAC-DRA) framework for optimizing smart electric vehicle (EV) charging ecosystems through the Smart2Charge application. The proposed system coordinates autonomous charging agents across networks of 250 EVs and 45 charging stations while adapting to dynamic environmental conditions through context-aware decision-making. Our multi-agent approach employs coordinated Deep Q\-Networks integrated with Graph Neural Networks and attention mechanisms, processing 20 contextual features including weather patterns, traffic conditions, grid load fluctuations, and electricity pricing.The framework balances five ecosystem stakeholders i.e. EV users (25\%), grid operators (20\%), charging station operators (20\%), fleet operators (20%), and environmental factors (15\%) through weighted coordination mechanisms and consensus protocols. Comprehensive validation using real-world datasets containing 441,077 charging transactions demonstrates superior performance compared to baseline algorithms including DDPG, A3C, PPO, and GNN approaches. The CAMAC\-DRA framework achieves 92\% coordination success rate, 15\% energy efficiency improvement, 10\% cost reduction, 20% grid strain decrease, and \2.3x faster convergence while maintaining 88\% training stability and 85\% sample efficiency. Real-world validation confirms commercial viability with Net Present Cost of -\$122,962 and 69\% cost reduction through renewable energy integration. The framework's unique contribution lies in developing context-aware multi-stakeholder coordination that successfully balances competing objectives while adapting to real-time variables, positioning it as a breakthrough solution for intelligent EV charging coordination and sustainable transportation electrification. --- ### Goal Reaching with Eikonal-Constrained Hierarchical Quasimetric Reinforcement Learning Authors: Vittorio Giammarino, Ahmed H. Qureshi Published: 2025-12-12 Categories: cs.LG, cs.RO, eess.SY, stat.ML arXiv: https://arxiv.org/abs/2512.12046v2 Goal-Conditioned Reinforcement Learning (GCRL) mitigates the difficulty of reward design by framing tasks as goal reaching rather than maximizing hand-crafted reward signals. In this setting, the optimal goal-conditioned value function naturally forms a quasimetric, motivating Quasimetric RL (QRL), which constrains value learning to quasimetric mappings and enforces local consistency through discrete, trajectory-based constraints. We propose Eikonal-Constrained Quasimetric RL (Eik-QRL), a continuous-time reformulation of QRL based on the Eikonal Partial Differential Equation (PDE). This PDE-based structure makes Eik-QRL trajectory-free, requiring only sampled states and goals, while improving out-of-distribution generalization. We provide theoretical guarantees for Eik-QRL and identify limitations that arise under complex dynamics. To address these challenges, we introduce Eik-Hierarchical QRL (Eik-HiQRL), which integrates Eik-QRL into a hierarchical decomposition. Empirically, Eik-HiQRL achieves state-of-the-art performance in offline goal-conditioned navigation and yields consistent gains over QRL in manipulation tasks, matching temporal-difference methods. --- ### AI as a Teaching Partner: Early Lessons from Classroom Codesign with Secondary Teachers Authors: Alex Liu, Lief Esbenshade, Shawon Sarkar, Zewei Tian, Min Sun, Zachary Zhang, Thomas Han, Yulia Lapicus, Kevin He Published: 2025-12-12 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2512.12045v1 This report presents a comprehensive account of the Colleague AI Classroom pilot, a collaborative design (co-design) study that brought generative AI technology directly into real classrooms. In this study, AI functioned as a third agent, an active participant that mediated feedback, supported inquiry, and extended teachers' instructional reach while preserving human judgment and teacher authority. Over seven weeks in spring 2025, 21 in-service teachers from four Washington State public school districts and one independent school integrated four AI-powered features of the Colleague AI Classroom into their instruction: Teaching Aide, Assessment and AI Grading, AI Tutor, and Student Growth Insights. More than 600 students in grades 6-12 used the platform in class at the direction of their teachers, who designed and facilitated the AI activities. During the Classroom pilot, teachers were co-design partners: they planned activities, implemented them with students, and provided weekly reflections on AI's role in classroom settings. The teachers' feedback guided iterative improvements for Colleague AI. The research team captured rich data through surveys, planning and reflection forms, group meetings, one-on-one interviews, and platform usage logs to understand where AI adds instructional value and where it requires refinement. --- ### Benchmarking Contextual Understanding for In-Car Conversational Systems Authors: Philipp Habicht, Lev Sorokin, Abdullah Saydemir, Ken E. Friedl, Andrea Stocco Published: 2025-12-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.12042v1 In-Car Conversational Question Answering (ConvQA) systems significantly enhance user experience by enabling seamless voice interactions. However, assessing their accuracy and reliability remains a challenge. This paper explores the use of Large Language Models (LLMs) alongside advanced prompting techniques and agent-based methods to evaluate the extent to which ConvQA system responses adhere to user utterances. The focus lies on contextual understanding and the ability to provide accurate venue recommendations considering user constraints and situational context. To evaluate utterance-response coherence using an LLM, we synthetically generate user utterances accompanied by correct and modified failure-containing system responses. We use input-output, chain-of-thought, self-consistency prompting, and multi-agent prompting techniques with 13 reasoning and non-reasoning LLMs of varying sizes and providers, including OpenAI, DeepSeek, Mistral AI, and Meta. We evaluate our approach on a case study involving restaurant recommendations. The most substantial improvements occur for small non-reasoning models when applying advanced prompting techniques, particularly multi-agent prompting. However, reasoning models consistently outperform non-reasoning models, with the best performance achieved using single-agent prompting with self-consistency. Notably, DeepSeek-R1 reaches an F1-score of 0.99 at a cost of 0.002 USD per request. Overall, the best balance between effectiveness and cost-time efficiency is reached with the non-reasoning model DeepSeek-V3. Our findings show that LLM-based evaluation offers a scalable and accurate alternative to traditional human evaluation for benchmarking contextual understanding in ConvQA systems. --- ### DFedReweighting: A Unified Framework for Objective-Oriented Reweighting in Decentralized Federated Learning Authors: Kaichuang Zhang, Wei Yin, Jinghao Yang, Ping Xu Published: 2025-12-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.12022v2 Decentralized federated learning (DFL) has emerged as a promising paradigm that enables multiple clients to collaboratively train machine learning models through iterative rounds of local training, communication, and aggregation, without relying on a central server. Nevertheless, DFL systems continue to face a range of challenges, including fairness and Byantine robustness. To address these challenges, we propose \textbf{DFedReweighting}, a unified aggregation framework that achieves diverse learning objectives in DFL via objective-oriented reweighting at the final step of each learning round. Specifically, for each client, the framework first evaluates a target performance metric (TPM) on a compact auxiliary dataset constructed from local data, yielding preliminary aggregation weights, which are subsequently refined by a customized reweighting strategy (CRS) to produce the final aggregation weights. Theoretically, we prove that an appropriate TPM-CRS combination guarantees linear convergence for general $L$-smoothand strongly convex functions. Empirical results consistently demonstrate that \textbf{DFedReweighting} significantly improves fairness and robustness against Byzantine attacks across diverse settings. Two multi-objective examples, spanning tasks across and within clients, further establish that a broad range of desired learning objectives can be accommodated by appropriately designing the TPM and CRS. Our code is available at https://github.com/KaichuangZhang/DFedReweighting. --- ### Exploring Spatial-Temporal Representation via Star Graph for mmWave Radar-based Human Activity Recognition Authors: Senhao Gao, Junqing Zhang, Luoyu Mei, Shuai Wang, Xuyu Wang Published: 2025-12-12 Categories: cs.CV, cs.LG, eess.IV arXiv: https://arxiv.org/abs/2512.12013v1 Human activity recognition (HAR) requires extracting accurate spatial-temporal features with human movements. A mmWave radar point cloud-based HAR system suffers from sparsity and variable-size problems due to the physical features of the mmWave signal. Existing works usually borrow the preprocessing algorithms for the vision-based systems with dense point clouds, which may not be optimal for mmWave radar systems. In this work, we proposed a graph representation with a discrete dynamic graph neural network (DDGNN) to explore the spatial-temporal representation of human movement-related features. Specifically, we designed a star graph to describe the high-dimensional relative relationship between a manually added static center point and the dynamic mmWave radar points in the same and consecutive frames. We then adopted DDGNN to learn the features residing in the star graph with variable sizes. Experimental results demonstrated that our approach outperformed other baseline methods using real-world HAR datasets. Our system achieved an overall classification accuracy of 94.27\%, which gets the near-optimal performance with a vision-based skeleton data accuracy of 97.25\%. We also conducted an inference test on Raspberry Pi~4 to demonstrate its effectiveness on resource-constraint platforms. \sh{ We provided a comprehensive ablation study for variable DDGNN structures to validate our model design. Our system also outperformed three recent radar-specific methods without requiring resampling or frame aggregators. --- ### Semantic-Drive: Democratizing Long-Tail Data Curation via Open-Vocabulary Grounding and Neuro-Symbolic VLM Consensus Authors: Antonio Guillen-Perez Published: 2025-12-12 Categories: cs.CV, cs.AI, cs.CL, cs.RO arXiv: https://arxiv.org/abs/2512.12012v2 The development of robust Autonomous Vehicles (AVs) is bottlenecked by the scarcity of "Long-Tail" training data. While fleets collect petabytes of video logs, identifying rare safety-critical events (e.g., erratic jaywalking, construction diversions) remains a manual, cost-prohibitive process. Existing solutions rely on coarse metadata search, which lacks precision, or cloud-based VLMs, which are privacy-invasive and expensive. We introduce Semantic-Drive, a local-first, neuro-symbolic framework for semantic data mining. Our approach decouples perception into two stages: (1) Symbolic Grounding via a real-time open-vocabulary detector (YOLOE) to anchor attention, and (2) Cognitive Analysis via a Reasoning VLM that performs forensic scene analysis. To mitigate hallucination, we implement a "System 2" inference-time alignment strategy, utilizing a multi-model "Judge-Scout" consensus mechanism. Benchmarked on the nuScenes dataset against the Waymo Open Dataset (WOD-E2E) taxonomy, Semantic-Drive achieves a Recall of 0.966 (vs. 0.475 for CLIP) and reduces Risk Assessment Error by 40% ccompared to the best single scout models. The system runs entirely on consumer hardware (NVIDIA RTX 3090), offering a privacy-preserving alternative to the cloud. --- ### Hold Onto That Thought: Assessing KV Cache Compression On Reasoning Authors: Minghui Liu, Aadi Palnitkar, Tahseen Rabbani, Hyunwoo Jae, Kyle Rui Sang, Dixi Yao, Shayan Shabihi, Fuheng Zhao, Tian Li, Ce Zhang, Furong Huang, Kunpeng Zhang Published: 2025-12-12 Categories: cs.CL, cs.AI, cs.PF arXiv: https://arxiv.org/abs/2512.12008v1 Large language models (LLMs) have demonstrated remarkable performance on long-context tasks, but are often bottlenecked by memory constraints. Namely, the KV cache, which is used to significantly speed up attention computations, grows linearly with context length. A suite of compression algorithms has been introduced to alleviate cache growth by evicting unimportant tokens. However, several popular strategies are targeted towards the prefill phase, i.e., processing long prompt context, and their performance is rarely assessed on reasoning tasks requiring long decoding. In particular, short but complex prompts, such as those in benchmarks like GSM8K and MATH500, often benefit from multi-step reasoning and self-reflection, resulting in thinking sequences thousands of tokens long. In this work, we benchmark the performance of several popular compression strategies on long-reasoning tasks. For the non-reasoning Llama-3.1-8B-Instruct, we determine that no singular strategy fits all, and that performance is heavily influenced by dataset type. However, we discover that H2O and our decoding-enabled variant of SnapKV are dominant strategies for reasoning models, indicating the utility of heavy-hitter tracking for reasoning traces. We also find that eviction strategies at low budgets can produce longer reasoning traces, revealing a tradeoff between cache size and inference costs. --- ### EnviroLLM: Resource Tracking and Optimization for Local AI Authors: Troy Allen Published: 2025-12-12 Categories: cs.LG, cs.CY arXiv: https://arxiv.org/abs/2512.12004v1 Large language models (LLMs) are increasingly deployed locally for privacy and accessibility, yet users lack tools to measure their resource usage, environmental impact, and efficiency metrics. This paper presents EnviroLLM, an open-source toolkit for tracking, benchmarking, and optimizing performance and energy consumption when running LLMs on personal devices. The system provides real-time process monitoring, benchmarking across multiple platforms (Ollama, LM Studio, vLLM, and OpenAI-compatible APIs), persistent storage with visualizations for longitudinal analysis, and personalized model and optimization recommendations. The system includes LLM-as-judge evaluations alongside energy and speed metrics, enabling users to assess quality-efficiency tradeoffs when testing models with custom prompts. --- ### Adversarial Attacks Against Deep Learning-Based Radio Frequency Fingerprint Identification Authors: Jie Ma, Junqing Zhang, Guanxiong Shen, Alan Marshall, Chip-Hong Chang Published: 2025-12-12 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2512.12002v1 Radio frequency fingerprint identification (RFFI) is an emerging technique for the lightweight authentication of wireless Internet of things (IoT) devices. RFFI exploits deep learning models to extract hardware impairments to uniquely identify wireless devices. Recent studies show deep learning-based RFFI is vulnerable to adversarial attacks. However, effective adversarial attacks against different types of RFFI classifiers have not yet been explored. In this paper, we carried out a comprehensive investigations into different adversarial attack methods on RFFI systems using various deep learning models. Three specific algorithms, fast gradient sign method (FGSM), projected gradient descent (PGD), and universal adversarial perturbation (UAP), were analyzed. The attacks were launched to LoRa-RFFI and the experimental results showed the generated perturbations were effective against convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and gated recurrent units (GRU). We further used UAP to launch practical attacks. Special factors were considered for the wireless context, including implementing real-time attacks, the effectiveness of the attacks over a period of time, etc. Our experimental evaluation demonstrated that UAP can successfully launch adversarial attacks against the RFFI, achieving a success rate of 81.7% when the adversary almost has no prior knowledge of the victim RFFI systems. --- ### Direct Confidence Alignment: Aligning Verbalized Confidence with Internal Confidence In Large Language Models Authors: Glenn Zhang, Treasure Mayowa, Jason Fan, Yicheng Fu, Aaron Sandoval, Sean O'Brien, Kevin Zhu Published: 2025-12-12 Categories: cs.CL arXiv: https://arxiv.org/abs/2512.11998v1 Producing trustworthy and reliable Large Language Models (LLMs) has become increasingly important as their usage becomes more widespread. Calibration seeks to achieve this by improving the alignment between the model's confidence and the actual likelihood of its responses being correct or desirable. However, it has been observed that the internal confidence of a model, derived from token probabilities, is not well aligned with its verbalized confidence, leading to misleading results with different calibration methods. In this paper, we propose Direct Confidence Alignment (DCA), a method using Direct Preference Optimization to align an LLM's verbalized confidence with its internal confidence rather than ground-truth accuracy, enhancing model transparency and reliability by ensuring closer alignment between the two confidence measures. We evaluate DCA across multiple open-weight LLMs on a wide range of datasets. To further assess this alignment, we also introduce three new calibration error-based metrics. Our results show that DCA improves alignment metrics on certain model architectures, reducing inconsistencies in a model's confidence expression. However, we also show that it can be ineffective on others, highlighting the need for more model-aware approaches in the pursuit of more interpretable and trustworthy LLMs. --- ### Log Anomaly Detection with Large Language Models via Knowledge-Enriched Fusion Authors: Anfeng Peng, Ajesh Koyatan Chathoth, Stephen Lee Published: 2025-12-12 Categories: cs.AI arXiv: https://arxiv.org/abs/2512.11997v1 System logs are a critical resource for monitoring and managing distributed systems, providing insights into failures and anomalous behavior. Traditional log analysis techniques, including template-based and sequence-driven approaches, often lose important semantic information or struggle with ambiguous log patterns. To address this, we present EnrichLog, a training-free, entry-based anomaly detection framework that enriches raw log entries with both corpus-specific and sample-specific knowledge. EnrichLog incorporates contextual information, including historical examples and reasoning derived from the corpus, to enable more accurate and interpretable anomaly detection. The framework leverages retrieval-augmented generation to integrate relevant contextual knowledge without requiring retraining. We evaluate EnrichLog on four large-scale system log benchmark datasets and compare it against five baseline methods. Our results show that EnrichLog consistently improves anomaly detection performance, effectively handles ambiguous log entries, and maintains efficient inference. Furthermore, incorporating both corpus- and sample-specific knowledge enhances model confidence and detection accuracy, making EnrichLog well-suited for practical deployments. --- ### V-REX: Benchmarking Exploratory Visual Reasoning via Chain-of-Questions Authors: Chenrui Fan, Yijun Liang, Shweta Bhardwaj, Kwesi Cobbina, Ming Li, Tianyi Zhou Published: 2025-12-12 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2512.11995v2 While many vision-language models (VLMs) are developed to answer well-defined, straightforward questions with highly specified targets, as in most benchmarks, they often struggle in practice with complex open-ended tasks, which usually require multiple rounds of exploration and reasoning in the visual space. Such visual thinking paths not only provide step-by-step exploration and verification as an AI detective but also produce better interpretations of the final answers. However, these paths are challenging to evaluate due to the large exploration space of intermediate steps. To bridge the gap, we develop an evaluation suite, ``Visual Reasoning with multi-step EXploration (V-REX)'', which is composed of a benchmark of challenging visual reasoning tasks requiring native multi-step exploration and an evaluation protocol. V-REX covers rich application scenarios across diverse domains. V-REX casts the multi-step exploratory reasoning into a Chain-of-Questions (CoQ) and disentangles VLMs' capability to (1) Planning: breaking down an open-ended task by selecting a chain of exploratory questions; and (2) Following: answering curated CoQ sequentially to collect information for deriving the final answer. By curating finite options of questions and answers per step, V-REX achieves a reliable quantitative and fine-grained analysis of the intermediate steps. By assessing SOTA proprietary and open-sourced VLMs, we reveal consistent scaling trends, significant differences between planning and following abilities, and substantial room for improvement in multi-step exploratory reasoning. --- ### Policy Gradient Algorithms for Age-of-Information Cost Minimization Authors: José-Ramón Vidal, Vicent Pla, Luis Guijarro, Israel Leyva-Mayorga Published: 2025-12-12 Categories: cs.NI, cs.IT, cs.LG arXiv: https://arxiv.org/abs/2512.11990v1 Recent developments in cyber-physical systems have increased the importance of maximizing the freshness of the information about the physical environment. However, optimizing the access policies of Internet of Things devices to maximize the data freshness, measured as a function of the Age-of-Information (AoI) metric, is a challenging task. This work introduces two algorithms to optimize the information update process in cyber-physical systems operating under the generate-at-will model, by finding an online policy without knowing the characteristics of the transmission delay or the age cost function. The optimization seeks to minimize the time-average cost, which integrates the AoI at the receiver and the data transmission cost, making the approach suitable for a broad range of scenarios. Both algorithms employ policy gradient methods within the framework of model-free reinforcement learning (RL) and are specifically designed to handle continuous state and action spaces. Each algorithm minimizes the cost using a distinct strategy for deciding when to send an information update. Moreover, we demonstrate that it is feasible to apply the two strategies simultaneously, leading to an additional reduction in cost. The results demonstrate that the proposed algorithms exhibit good convergence properties and achieve a time-average cost within 3% of the optimal value, when the latter is computable. A comparison with other state-of-the-art methods shows that the proposed algorithms outperform them in one or more of the following aspects: being applicable to a broader range of scenarios, achieving a lower time-average cost, and requiring a computational cost at least one order of magnitude lower. --- ### CARI4D: Category Agnostic 4D Reconstruction of Human-Object Interaction Authors: Xianghui Xie, Bowen Wen, Yan Chang, Hesam Rabeti, Jiefeng Li, Ye Yuan, Gerard Pons-Moll, Stan Birchfield Published: 2025-12-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.11988v3 Accurate capture of human-object interaction from ubiquitous sensors like RGB cameras is important for applications in human understanding, gaming, and robot learning. However, inferring 4D interactions from a single RGB view is highly challenging due to the unknown object and human information, depth ambiguity, occlusion, and complex motion, which hinder consistent 3D and temporal reconstruction. Previous methods simplify the setup by assuming ground truth object template or constraining to a limited set of object categories. We present CARI4D, the first category-agnostic method that reconstructs spatially and temporarily consistent 4D human-object interaction at metric scale from monocular RGB videos. To this end, we propose a pose hypothesis selection algorithm that robustly integrates the individual predictions from foundation models, jointly refine them through a learned render-and-compare paradigm to ensure spatial, temporal and pixel alignment, and finally reasoning about intricate contacts for further refinement satisfying physical constraints. Experiments show that our method outperforms prior art by 38% on in-distribution dataset and 36% on unseen dataset in terms of reconstruction error. Our model generalizes beyond the training categories and thus can be applied zero-shot to in-the-wild internet videos. Our code and pretrained models will be publicly released. --- ### Learning to Extract Context for Context-Aware LLM Inference Authors: Minseon Kim, Lucas Caccia, Zhengyan Shi, Matheus Pereira, Marc-Alexandre Côté, Xingdi Yuan, Alessandro Sordoni Published: 2025-12-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.11986v1 User prompts to large language models (LLMs) are often ambiguous or under-specified, and subtle contextual cues shaped by user intentions, prior knowledge, and risk factors strongly influence what constitutes an appropriate response. Misinterpreting intent or risks may lead to unsafe outputs, while overly cautious interpretations can cause unnecessary refusal of benign requests. In this paper, we question the conventional framework in which LLMs generate immediate responses to requests without considering broader contextual factors. User requests are situated within broader contexts such as intentions, knowledge, and prior experience, which strongly influence what constitutes an appropriate answer. We propose a framework that extracts and leverages such contextual information from the user prompt itself. Specifically, a reinforcement learning based context generator, designed in an autoencoder-like fashion, is trained to infer contextual signals grounded in the prompt and use them to guide response generation. This approach is particularly important for safety tasks, where ambiguous requests may bypass safeguards while benign but confusing requests can trigger unnecessary refusals. Experiments show that our method reduces harmful responses by an average of 5.6% on the SafetyInstruct dataset across multiple foundation models and improves the harmonic mean of attack success rate and compliance on benign prompts by 6.2% on XSTest and WildJailbreak. These results demonstrate the effectiveness of context extraction for safer and more reliable LLM inferences. --- ### Evidence-Driven Decision Support for AI Model Selection in Research Software Engineering Authors: Alireza Joonbakhsh, Alireza Rostami, AmirMohammad Kamalinia, Ali Nazeri, Farshad Khunjush, Bedir Tekinerdogan, Siamak Farshidi Published: 2025-12-12 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2512.11984v1 The rapid proliferation of artificial intelligence (AI) models and methods presents growing challenges for research software engineers and researchers who must select, integrate, and maintain appropriate models within complex research workflows. Model selection is often performed in an ad hoc manner, relying on fragmented metadata and individual expertise, which can undermine reproducibility, transparency, and overall research software quality. This work proposes a structured and evidence-driven approach to support AI model selection that aligns with both technical and contextual requirements. We conceptualize AI model selection as a Multi-Criteria Decision-Making (MCDM) problem and introduce an evidence-based decision-support framework that integrates automated data collection pipelines, a structured knowledge graph, and MCDM principles. Following the Design Science Research methodology, the proposed framework (ModelSelect) is empirically validated through 50 real-world case studies and comparative experiments against leading generative AI systems. The evaluation results show that ModelSelect produces reliable, interpretable, and reproducible recommendations that closely align with expert reasoning. Across the case studies, the framework achieved high coverage and strong rationale alignment in both model and library recommendation tasks, performing comparably to generative AI assistants while offering superior traceability and consistency. By framing AI model selection as an MCDM problem, this work establishes a rigorous foundation for transparent and reproducible decision support in research software engineering. The proposed framework provides a scalable and explainable pathway for integrating empirical evidence into AI model recommendation processes, ultimately improving the quality and robustness of research software decision-making. --- ### Semantic search for 100M+ galaxy images using AI-generated captions Authors: Nolan Koblischke, Liam Parker, Francois Lanusse, Jo Bovy, Irina Espejo, Shirley Ho Published: 2025-12-12 Categories: astro-ph.IM, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2512.11982v2 Finding scientifically interesting phenomena through slow manual labeling campaigns severely limits our ability to explore the billions of galaxy images produced by telescopes. In this work, we develop a pipeline to create a semantic search engine from completely unlabeled image data. Our method leverages Vision-Language Models (VLMs) to generate descriptions for galaxy images, then contrastively aligns a pre-trained astronomy foundation model with these embedded descriptions to produce searchable embeddings at scale. We find that current VLMs provide descriptions that are sufficiently informative to train a semantic search model that outperforms direct image similarity search. Our model, AION-Search, achieves state-of-the-art zero-shot performance on finding rare phenomena despite training on randomly selected images with no deliberate curation for rare cases. Furthermore, we introduce a VLM-based re-ranking method that nearly doubles the recall for our most challenging targets in the top-100 results. For the first time, AION-Search enables flexible semantic search for over 100 million galaxy images, enabling discovery from previously infeasible searches, including the identification of 36 new extragalactic stellar stream candidates. More broadly, our work provides an approach for making large, unlabeled scientific image archives semantically searchable, expanding data exploration capabilities in fields from Earth observation to microscopy. The code, data, and app are publicly available at https://github.com/NolanKoblischke/AION-Search --- ### Designing The Internet of Agents: A Framework for Trustworthy, Transparent, and Collaborative Human-Agent Interaction (HAX) Authors: Marc Scibelli, Krystelle Gonzalez Papaux, Julia Valenti, Srishti Kush Published: 2025-12-12 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2512.11979v1 The rise of generative and autonomous agents marks a fundamental shift in computing, demanding a rethinking of how humans collaborate with probabilistic, partially autonomous systems. We present the Human-AI-Experience (HAX) framework, a comprehensive, three-phase approach that establishes design foundations for trustworthy, transparent, and collaborative agentic interaction. HAX integrates behavioral heuristics, a schema-driven SDK enforcing structured and safe outputs, and a behavioral proxy concept that orchestrates agent activity to reduce cognitive load. A validated catalog of mixed-initiative design patterns further enables intent preview, iterative alignment, trust repair, and multi-agent narrative coherence. Grounded in Time, Interaction, and Performance (TIP) theory, HAX reframes multi-agent systems as colleagues, offering the first end-to-end framework that bridges trust theory, interface design, and infrastructure for the emerging Internet of Agents. --- ### A Comparative Analysis of Semiconductor Wafer Map Defect Detection with Image Transformer Authors: Sushmita Nath Published: 2025-12-12 Categories: cs.CV arXiv: https://arxiv.org/abs/2512.11977v1 Predictive maintenance is an important sector in modern industries which improves fault detection and cost reduction processes. By using machine learning algorithms in the whole process, the defects detection process can be implemented smoothly. Semiconductor is a sensitive maintenance field that requires predictability in work. While convolutional neural networks (CNNs) such as VGG-19, Xception and Squeeze-Net have demonstrated solid performance in image classification for semiconductor wafer industry, their effectiveness often declines in scenarios with limited and imbalanced data. This study investigates the use of the Data-Efficient Image Transformer (DeiT) for classifying wafer map defects under data-constrained conditions. Experimental results reveal that the DeiT model achieves highest classification accuracy of 90.83%, outperforming CNN models such as VGG-19(65%), SqueezeNet(82%), Xception(66%) and Hybrid(67%). DeiT also demonstrated superior F1-score (90.78%) and faster training convergence, with enhanced robustness in detecting minority defect classes. These findings highlight the potential of transformer-based models like DeiT in semiconductor wafer defect detection and support predictive maintenance strategies within semiconductor fabrication processes. --- ### Pre-training vision models for the classification of alerts from wide-field time-domain surveys Authors: Nabeel Rehemtulla, Adam A. Miller, Mike Walmsley, Ved G. Shah, Theophile Jegou du Laz, Michael W. Coughlin, Argyro Sasli, Joshua Bloom, Christoffer Fremling, Matthew J. Graham, Steven L. Groom, David Hale et al. Published: 2025-12-12 Categories: astro-ph.IM, cs.CV arXiv: https://arxiv.org/abs/2512.11957v2 Modern wide-field time-domain surveys facilitate the study of transient, variable and moving phenomena by conducting image differencing and relaying alerts to their communities. Machine learning tools have been used on data from these surveys and their precursors for more than a decade, and convolutional neural networks (CNNs), which make predictions directly from input images, saw particularly broad adoption through the 2010s. Since then, continually rapid advances in computer vision have transformed the standard practices around using such models. It is now commonplace to use standardized architectures pre-trained on large corpora of everyday images (e.g., ImageNet). In contrast, time-domain astronomy studies still typically design custom CNN architectures and train them from scratch. Here, we explore the effects of adopting various pre-training regimens and standardized model architectures on the performance of alert classification. We find that the resulting models match or outperform a custom, specialized CNN like what is typically used for filtering alerts. Moreover, our results show that pre-training on galaxy images from Galaxy Zoo tends to yield better performance than pre-training on ImageNet or training from scratch. We observe that the design of standardized architectures are much better optimized than the custom CNN baseline, requiring significantly less time and memory for inference despite having more trainable parameters. On the eve of the Legacy Survey of Space and Time and other image-differencing surveys, these findings advocate for a paradigm shift in the creation of vision models for alerts, demonstrating that greater performance and efficiency, in time and in data, can be achieved by adopting the latest practices from the computer vision field. --- ### Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors Authors: Max McGuinness, Alex Serrano, Luke Bailey, Scott Emmons Published: 2025-12-12 Categories: cs.LG arXiv: https://arxiv.org/abs/2512.11949v1 Activation monitoring, which probes a model's internal states using lightweight classifiers, is an emerging tool for AI safety. However, its worst-case robustness under a misalignment threat model--where a model might learn to actively conceal its internal states--remains untested. Focusing on this threat model, we ask: could a model learn to evade previously unseen activation monitors? Our core contribution is to stress-test the learnability of this behavior. We demonstrate that finetuning can create Neural Chameleons: models capable of zero-shot evading activation monitors. Specifically, we fine-tune an LLM to evade monitors for a set of benign concepts (e.g., languages, HTML) when conditioned on a trigger of the form: "You are being probed for {concept}". We show that this learned mechanism generalizes zero-shot: by substituting {concept} with a safety-relevant term like 'deception', the model successfully evades previously unseen safety monitors. We validate this phenomenon across diverse model families (Llama, Gemma, Qwen), showing that the evasion succeeds even against monitors trained post hoc on the model's frozen weights. This evasion is highly selective, targeting only the specific concept mentioned in the trigger, and having a modest impact on model capabilities on standard benchmarks. Using Gemma-2-9b-it as a case study, a mechanistic analysis reveals this is achieved via a targeted manipulation that moves activations into a low-dimensional subspace. While stronger defenses like monitor ensembles and non-linear classifiers show greater resilience, the model retains a non-trivial evasion capability. Our work provides a proof-of-concept for this failure mode and a tool to evaluate the worst-case robustness of monitoring techniques against misalignment threat models. --- ### Global Sensitivity Analysis for Engineering Design Based on Individual Conditional Expectations Authors: Pramudita Satria Palar, Paul Saves, Rommel G. Regis, Koji Shimoyama, Shigeru Obayashi, Nicolas Verstaevel, Joseph Morlier Published: 2025-12-12 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2512.11946v3 Explainable machine learning techniques have gained increasing attention in engineering applications, especially in aerospace design and analysis, where understanding how input variables influence data-driven models is essential. Partial Dependence Plots (PDPs) are widely used for interpreting black-box models by showing the average effect of an input variable on the prediction. However, their global sensitivity metric can be misleading when strong interactions are present, as averaging tends to obscure interaction effects. To address this limitation, we propose a global sensitivity metric based on Individual Conditional Expectation (ICE) curves. The method computes the expected feature importance across ICE curves, along with their standard deviation, to more effectively capture the influence of interactions. We provide a mathematical proof demonstrating that the PDP-based sensitivity is a lower bound of the proposed ICE-based metric under truncated orthogonal polynomial expansion. In addition, we introduce an ICE-based correlation value to quantify how interactions modify the relationship between inputs and the output. Comparative evaluations were performed on three cases: a 5-variable analytical function, a 5-variable wind-turbine fatigue problem, and a 9-variable airfoil aerodynamics case, where ICE-based sensitivity was benchmarked against PDP, SHapley Additive exPlanations (SHAP), and Sobol' indices. The results show that ICE-based feature importance provides richer insights than the traditional PDP-based approach, while visual interpretations from PDP, ICE, and SHAP complement one another by offering multiple perspectives. --- ### Interval Fisher's Discriminant Analysis and Visualisation Authors: Diogo Pinheiro, M. Rosário Oliveira, Igor Kravchenko, Lina Oliveira Published: 2025-12-12 Categories: stat.ML, cs.LG, math.ST arXiv: https://arxiv.org/abs/2512.11945v1 In Data Science, entities are typically represented by single valued measurements. Symbolic Data Analysis extends this framework to more complex structures, such as intervals and histograms, that express internal variability. We propose an extension of multiclass Fisher's Discriminant Analysis to interval-valued data, using Moore's interval arithmetic and the Mallows' distance. Fisher's objective function is generalised to consider simultaneously the contributions of the centres and the ranges of intervals and is numerically maximised. The resulting discriminant directions are then used to classify interval-valued observations.To support visual assessment, we adapt the class map, originally introduced for conventional data, to classifiers that assign labels through minimum distance rules. We also extend the silhouette plot to this setting and use stacked mosaic plots to complement the visual display of class assignments. Together, these graphical tools provide insight into classifier performance and the strength of class membership. Applications to real datasets illustrate the proposed methodology and demonstrate its value in interpreting classification results for interval-valued data. --- ### A Review of Learning-Based Motion Planning: Toward a Data-Driven Optimal Control Approach Authors: Jia Hu, Yang Chang, Haoran Wang Published: 2025-12-12 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2512.11944v2 Motion planning for autonomous driving (AD) faces a critical trade-off. While traditional rule-based pipelines offer verifiable safety and interpretability, they often fail to generalize in complex scenarios. Conversely, emerging learning-based methods-including imitation learning (IL), reinforcement learning (RL), and generative AI-offer greater adaptability but are often constrained by opacity and safety risks. Existing surveys typically analyze these AI methods in isolation, overlooking the potential of integrating them with rigorous control frameworks. To bridge this gap, this paper presents the first systematic review of the Data-Driven Optimal Control (DDOC) paradigm, explicitly examining how it synergizes the theoretical guarantees of optimal control with the adaptive capabilities of modern machine learning. Building on this framework, we propose the first roadmap for DDOC-based motion planning, structuring its implementation into three critical dimensions: customization, dynamics adaptation, and self-tuning. Finally, to close the remaining reality gap, we identify four future research directions, thereby accelerating the transition to trustworthy and human-like autonomous driving. --- ### How AI Agents Follow the Herd of AI? Network Effects, History, and Machine Optimism Authors: Yu Liu, Wenwen Li, Yifan Dou, Guangnan Ye Published: 2025-12-12 Categories: cs.MA, cs.AI, econ.GN arXiv: https://arxiv.org/abs/2512.11943v1 Understanding decision-making in multi-AI-agent frameworks is crucial for analyzing strategic interactions in network-effect-driven contexts. This study investigates how AI agents navigate network-effect games, where individual payoffs depend on peer participatio--a context underexplored in multi-agent systems despite its real-world prevalence. We introduce a novel workflow design using large language model (LLM)-based agents in repeated decision-making scenarios, systematically manipulating price trajectories (fixed, ascending, descending, random) and network-effect strength. Our key findings include: First, without historical data, agents fail to infer equilibrium. Second, ordered historical sequences (e.g., escalating prices) enable partial convergence under weak network effects but strong effects trigger persistent "AI optimism"--agents overestimate participation despite contradictory evidence. Third, randomized history disrupts convergence entirely, demonstrating that temporal coherence in data shapes LLMs' reasoning, unlike humans. These results highlight a paradigm shift: in AI-mediated systems, equilibrium outcomes depend not just on incentives, but on how history is curated, which is impossible for human. --- ### Hypergame Rationalisability: Solving Agent Misalignment In Strategic Play Authors: Vince Trencsenyi Published: 2025-12-12 Categories: cs.AI, cs.FL, cs.GT arXiv: https://arxiv.org/abs/2512.11942v1 Differences in perception, information asymmetries, and bounded rationality lead game-theoretic players to derive a private, subjective view of the game that may diverge from the underlying ground-truth scenario and may be misaligned with other players' interpretations. While typical game-theoretic assumptions often overlook such heterogeneity, hypergame theory provides the mathematical framework to reason about mismatched mental models. Although hypergames have recently gained traction in dynamic applications concerning uncertainty, their practical adoption in multi-agent system research has been hindered by the lack of a unifying, formal, and practical representation language, as well as scalable algorithms for managing complex hypergame structures and equilibria. Our work addresses this gap by introducing a declarative, logic-based domain-specific language for encoding hypergame structures and hypergame solution concepts. Leveraging answer-set programming, we develop an automated pipeline for instantiating hypergame structures and running our novel hypergame rationalisation procedure, a mechanism for finding belief structures that justify seemingly irrational outcomes. The proposed language establishes a unifying formalism for hypergames and serves as a foundation for developing nuanced, belief-based heterogeneous reasoners, offering a verifiable context with logical guarantees. Together, these contributions establish the connection between hypergame theory, multi-agent systems, and strategic AI. --- ### DynaPURLS: Dynamic Refinement of Part-Aware Representations for Skeleton-Based Zero-Shot Action Recognition Authors: Jingmin Zhu, Anqi Zhu, James Bailey, Jun Liu, Hossein Rahmani, Mohammed Bennamoun, Farid Boussaid, Qiuhong Ke Published: 2025-12-12 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2512.11941v2 Zero-shot skeleton-based action recognition (ZS-SAR) is fundamentally constrained by prevailing approaches that rely on aligning skeleton features with static, class-level semantics. This coarse-grained alignment fails to bridge the domain shift between seen and unseen classes, thereby impeding the effective transfer of fine-grained visual knowledge. To address these limitations, we introduce \textbf{DynaPURLS}, a unified framework that establishes robust, multi-scale visual-semantic correspondences and dynamically refines them at inference time to enhance generalization. Our framework leverages a large language model to generate hierarchical textual descriptions that encompass both global movements and local body-part dynamics. Concurrently, an adaptive partitioning module produces fine-grained visual representations by semantically grouping skeleton joints. To fortify this fine-grained alignment against the train-test domain shift, DynaPURLS incorporates a dynamic refinement module. During inference, this module adapts textual features to the incoming visual stream via a lightweight learnable projection. This refinement process is stabilized by a confidence-aware, class-balanced memory bank, which mitigates error propagation from noisy pseudo-labels. Extensive experiments on three large-scale benchmark datasets, including NTU RGB+D 60/120 and PKU-MMD, demonstrate that DynaPURLS significantly outperforms prior art, setting new state-of-the-art records. The source code is made publicly available at https://github.com/Alchemist0754/DynaPURLS --- ### Contextual Peano Scan and Fast Image Segmentation Using Hidden and Evidential Markov Chains Authors: Clément Fernandes, Wojciech Pieczynski Published: 2025-12-12 Categories: cs.CV, math.ST, stat.AP arXiv: https://arxiv.org/abs/2512.11939v1 Transforming bi-dimensional sets of image pixels into mono-dimensional sequences with a Peano scan (PS) is an established technique enabling the use of hidden Markov chains (HMCs) for unsupervised image segmentation. Related Bayesian segmentation methods can compete with hidden Markov fields (HMFs)-based ones and are much faster. PS has recently been extended to the contextual PS, and some initial experiments have shown the value of the associated HMC model, denoted as HMC-CPS, in image segmentation. Moreover, HMCs have been extended to hidden evidential Markov chains (HEMCs), which are capable of improving HMC-based Bayesian segmentation. In this study, we introduce a new HEMC-CPS model by simultaneously considering contextual PS and evidential HMC. We show its effectiveness for Bayesian maximum posterior mode (MPM) segmentation using synthetic and real images. Segmentation is performed in an unsupervised manner, with parameters being estimated using the stochastic expectation--maximization (SEM) method. The new HEMC-CPS model presents potential for the modeling and segmentation of more complex images, such as three-dimensional or multi-sensor multi-resolution images. Finally, the HMC-CPS and HEMC-CPS models are not limited to image segmentation and could be used for any kind of spatially correlated data. ---