# Research Papers — 2025-12 · bundle page 152 of 175

> Full-text markdown of papers #7551–7600 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/152
[← Previous bundle](https://aipod.dev/papers/archive/2025-12/page/151.md)
[Next bundle →](https://aipod.dev/papers/archive/2025-12/page/153.md)

### Cognitive Mirrors: Exploring the Diverse Functional Roles of Attention Heads in LLM Reasoning

Authors: Xueqi Ma, Jun Wang, Yanbei Jiang, Sarah Monazam Erfani, Tongliang Liu, James Bailey
Published: 2025-12-03
Categories: q-bio.NC, cs.AI
arXiv: https://arxiv.org/abs/2512.10978v1

Large language models (LLMs) have achieved state-of-the-art performance in a variety of tasks, but remain largely opaque in terms of their internal mechanisms. Understanding these mechanisms is crucial to improve their reasoning abilities. Drawing inspiration from the interplay between neural processes and human cognition, we propose a novel interpretability framework to systematically analyze the roles and behaviors of attention heads, which are key components of LLMs. We introduce CogQA, a dataset that decomposes complex questions into step-by-step subquestions with a chain-of-thought design, each associated with specific cognitive functions such as retrieval or logical reasoning. By applying a multi-class probing method, we identify the attention heads responsible for these functions. Our analysis across multiple LLM families reveals that attention heads exhibit functional specialization, characterized as cognitive heads. These cognitive heads exhibit several key properties: they are universally sparse, vary in number and distribution across different cognitive functions, and display interactive and hierarchical structures. We further show that cognitive heads play a vital role in reasoning tasks - removing them leads to performance degradation, while augmenting them enhances reasoning accuracy. These insights offer a deeper understanding of LLM reasoning and suggest important implications for model design, training, and fine-tuning strategies.

---

### Peek-a-Boo Reasoning: Contrastive Region Masking in MLLMs

Authors: Isha Chaturvedi, Anjana Nair, Yushen Li, Adhitya Rajendra Kumar, Kevin Zhu, Sunishchal Dev, Ashwinee Panda, Vasu Sharma
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.08976v1

We introduce Contrastive Region Masking (CRM), a training free diagnostic that reveals how multimodal large language models (MLLMs) depend on specific visual regions at each step of chain-of-thought (CoT) reasoning. Unlike prior approaches limited to final answers or attention maps, CRM provides causal, step-level attribution by systematically masking annotated regions and contrasting the resulting reasoning traces with unmasked baselines. Applied to datasets such as VisArgs, CRM reveals distinct failure modes: some models preserve reasoning structure, but hallucinate when evidence is missing, while others ground tightly to visual cues yet collapse under perturbations. By shifting the evaluation from correctness of answers to faithfulness of reasoning, CRM reframes visual benchmarks as diagnostic tools, highlighting the need for multimodal evaluation frameworks that measure not just performance, but also robustness and fidelity of reasoning.

---

### FuXi-Nowcast: Environment-conditioned deep learning for severe convection nowcasting

Authors: Lei Chen, Zijian Zhu, Xiaoran Zhuang, Tianyuan Qi, Yuxuan Feng, Xiaohui Zhong, Hao Li
Published: 2025-12-03
Categories: physics.ao-ph, cs.LG
arXiv: https://arxiv.org/abs/2512.08974v2

Severe convection produces localized hazards that often require warnings before radar echoes fully reveal storm development. Convective initiation and the maintenance of intense convection remain challenging for radar-only nowcasting because pre-convective signals may be absent from recent radar observations and strong echoes often decay rapidly in forecasts. Here we present FuXi-Nowcast, an environment-conditioned deep learning system that combines high-resolution observations with three-dimensional atmospheric forecasts to predict composite reflectivity, precipitation, wind gusts, and surface variables up to 12 h ahead. In April--July 2024 evaluations over East China, FuXi-Nowcast outperforms operational numerical, persistence and extrapolation baselines for reflectivity and precipitation. Case studies, diagnostics, and ablation experiments suggest that atmospheric moisture information and explicit preservation of strong convective signals contribute to forecasts of convective initiation and maintenance. These results show that environmental conditioning can mitigate important failure modes of radar-only nowcasting for high-impact convective weather.

---

### Benchmarking CXR Foundation Models With Publicly Available MIMIC-CXR and NIH-CXR14 Datasets

Authors: Jiho Shin, Dominic Marshall, Matthieu Komorowski
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.06014v1

Recent foundation models have demonstrated strong performance in medical image representation learning, yet their comparative behaviour across datasets remains underexplored. This work benchmarks two large-scale chest X-ray (CXR) embedding models (CXR-Foundation (ELIXR v2.0) and MedImagelnsight) on public MIMIC-CR and NIH ChestX-ray14 datasets. Each model was evaluated using a unified preprocessing pipeline and fixed downstream classifiers to ensure reproducible comparison. We extracted embeddings directly from pre-trained encoders, trained lightweight LightGBM classifiers on multiple disease labels, and reported mean AUROC, and F1-score with 95% confidence intervals. MedImageInsight achieved slightly higher performance across most tasks, while CXR-Foundation exhibited strong cross-dataset stability. Unsupervised clustering of MedImageIn-sight embeddings further revealed a coherent disease-specific structure consistent with quantitative results. The results highlight the need for standardised evaluation of medical foundation models and establish reproducible baselines for future multimodal and clinical integration studies.

---

### VAT: Vision Action Transformer by Unlocking Full Representation of ViT

Authors: Wenhao Li, Chengwei Ma, Weixin Mao
Published: 2025-12-03
Categories: cs.CV, cs.RO
arXiv: https://arxiv.org/abs/2512.06013v2

In robot learning, Vision Transformers (ViTs) are standard for visual perception, yet most methods discard valuable information by using only the final layer's features. We argue this provides an insufficient representation and propose the Vision Action Transformer (VAT), a novel architecture that is extended from ViT and unlocks the full feature hierarchy of ViT. VAT processes specialized action tokens with visual features across all transformer layers, enabling a deep and progressive fusion of perception and action generation. On a suite of simulated manipulation tasks, VAT achieves a 98.15\% average success rate across four LIBERO benchmarks, establishing a new state-of-the-art by outperforming prior methods like OpenVLA-OFT. Our work presents not only a powerful model for imitation learning but also demonstrates the critical importance of leveraging the complete ''representation trajectory'' of vision models to advance robotic policy. The GitHub URL for the project code is https://github.com/sellerbubble/VAT.

---

### High-Throughput Unsupervised Profiling of the Morphology of 316L Powder Particles for Use in Additive Manufacturing

Authors: Emmanuel Akeweje, Conall Kirk, Chi-Wai Chan, Denis Dowling, Mimi Zhang
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.06012v2

Selective Laser Melting (SLM) is a powder-bed additive manufacturing technique whose part quality depends critically on feedstock morphology. However, conventional powder characterization methods are low-throughput and qualitative, failing to capture the heterogeneity of industrial-scale batches. We present an automated, machine learning framework that couples high-throughput imaging with shape extraction and clustering to profile metallic powder morphology at scale. We develop and evaluate three clustering pipelines: an autoencoder pipeline, a shape-descriptor pipeline, and a functional-data pipeline. Across a dataset of approximately 126,000 powder images (0.5-102 micrometer diameter), internal validity metrics identify the Fourier-descriptor + k-means pipeline as the most effective, achieving the lowest Davies-Bouldin index and highest Calinski-Harabasz score while maintaining sub-millisecond runtime per particle on a standard desktop workstation. Although the present work focuses on establishing the morphological-clustering framework, the resulting shape groups form a basis for future studies examining their relationship to flowability, packing density, and SLM part quality. Overall, this unsupervised learning framework enables rapid, automated assessment of powder morphology and supports tracking of shape evolution across reuse cycles, offering a path toward real-time feedstock monitoring in SLM workflows.

---

### Fast and Flexible Robustness Certificates for Semantic Segmentation

Authors: Thomas Massena, Corentin Friedrich, Franck Mamalet, Mathieu Serrurier
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.06010v1

Deep Neural Networks are vulnerable to small perturbations that can drastically alter their predictions for perceptually unchanged inputs. The literature on adversarially robust Deep Learning attempts to either enhance the robustness of neural networks (e.g, via adversarial training) or to certify their decisions up to a given robustness level (e.g, by using randomized smoothing, formal methods or Lipschitz bounds). These studies mostly focus on classification tasks and few efficient certification procedures currently exist for semantic segmentation. In this work, we introduce a new class of certifiably robust Semantic Segmentation networks with built-in Lipschitz constraints that are efficiently trainable and achieve competitive pixel accuracy on challenging datasets such as Cityscapes. Additionally, we provide a novel framework that generalizes robustness certificates for semantic segmentation tasks, where we showcase the flexibility and computational efficiency of using Lipschitz networks. Our approach unlocks real-time compatible certifiably robust semantic segmentation for the first time. Moreover, it allows the computation of worst-case performance under $\ell_2$ attacks of radius $ε$ across a wide range of performance measures. Crucially, we benchmark the runtime of our certification process and find our approach to be around 600 times faster than randomized smoothing methods at inference with comparable certificates on an NVIDIA A100 GPU. Finally, we evaluate the tightness of our worstcase certificates against state-of-the-art adversarial attacks to further validate the performance of our method.

---

### Semantic Temporal Single-photon LiDAR

Authors: Fang Li, Tonglin Mu, Shuling Li, Junran Guo, Keyuan Li, Jianing Li, Ziyang Luo, Xiaodong Fan, Ye Chen, Yunfeng Liu, Hong Cai, Lip Ket Chin et al.
Published: 2025-12-03
Categories: eess.IV, cs.CV, quant-ph
arXiv: https://arxiv.org/abs/2512.06008v1

Temporal single-photon (TSP-) LiDAR presents a promising solution for imaging-free target recognition over long distances with reduced size, cost, and power consumption. However, existing TSP-LiDAR approaches are ineffective in handling open-set scenarios where unknown targets emerge, and they suffer significant performance degradation under low signal-to-noise ratio (SNR) and short acquisition times (fewer photons). Here, inspired by semantic communication, we propose a semantic TSP-LiDAR based on a self-updating semantic knowledge base (SKB), in which the target recognition processing of TSP-LiDAR is formulated as a semantic communication. The results, both simulation and experiment, demonstrate that our approach surpasses conventional methods, particularly under challenging conditions of low SNR and limited acquisition time. More importantly, our self-updating SKB mechanism can dynamically update the semantic features of newly encountered targets in the SKB, enabling continuous adaptation without the need for extensive retraining of the neural network. In fact, a recognition accuracy of 89% is achieved on nine types of unknown targets in real-world experiments, compared to 66% without the updating mechanism. These findings highlight the potential of our framework for adaptive and robust target recognition in complex and dynamic environments.

---

### EFDiT: Efficient Fine-grained Image Generation Using Diffusion Transformer Models

Authors: Kun Wang, Donglin Di, Tonghua Su, Lei Fan
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.05152v1

Diffusion models are highly regarded for their controllability and the diversity of images they generate. However, class-conditional generation methods based on diffusion models often focus on more common categories. In large-scale fine-grained image generation, issues of semantic information entanglement and insufficient detail in the generated images still persist. This paper attempts to introduce a concept of a tiered embedder in fine-grained image generation, which integrates semantic information from both super and child classes, allowing the diffusion model to better incorporate semantic information and address the issue of semantic entanglement. To address the issue of insufficient detail in fine-grained images, we introduce the concept of super-resolution during the perceptual information generation stage, enhancing the detailed features of fine-grained images through enhancement and degradation models. Furthermore, we propose an efficient ProAttention mechanism that can be effectively implemented in the diffusion model. We evaluate our method through extensive experiments on public benchmarks, demonstrating that our approach outperforms other state-of-the-art fine-tuning methods in terms of performance.

---

### TwinFlow: Realizing One-step Generation on Large Models with Self-adversarial Flows

Authors: Zhenglin Cheng, Peng Sun, Jianguo Li, Tao Lin
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.05150v2

Recent advances in large multi-modal generative models have demonstrated impressive capabilities in multi-modal generation, including image and video generation. These models are typically built upon multi-step frameworks like diffusion and flow matching, which inherently limits their inference efficiency (requiring 40-100 Number of Function Evaluations (NFEs)). While various few-step methods aim to accelerate the inference, existing solutions have clear limitations. Prominent distillation-based methods, such as progressive and consistency distillation, either require an iterative distillation procedure or show significant degradation at very few steps (< 4-NFE). Meanwhile, integrating adversarial training into distillation (e.g., DMD/DMD2 and SANA-Sprint) to enhance performance introduces training instability, added complexity, and high GPU memory overhead due to the auxiliary trained models. To this end, we propose TwinFlow, a simple yet effective framework for training 1-step generative models that bypasses the need of fixed pretrained teacher models and avoids standard adversarial networks during training, making it ideal for building large-scale, efficient models. On text-to-image tasks, our method achieves a GenEval score of 0.83 in 1-NFE, outperforming strong baselines like SANA-Sprint (a GAN loss-based framework) and RCGM (a consistency-based framework). Notably, we demonstrate the scalability of TwinFlow by full-parameter training on Qwen-Image-20B and transform it into an efficient few-step generator. With just 1-NFE, our approach matches the performance of the original 100-NFE model on both the GenEval and DPG-Bench benchmarks, reducing computational cost by $100\times$ with minor quality degradation. Project page is available at https://zhenglin-cheng.com/twinflow.

---

### A Conceptual Model for AI Adoption in Financial Decision-Making: Addressing the Unique Challenges of Small and Medium-Sized Enterprises

Authors: Manh Chien Vu, Thang Le Dinh, Manh Chien Vu, Tran Duc Le, Thi Lien Huong Nguyen
Published: 2025-12-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2512.04339v1

The adoption of artificial intelligence (AI) offers transformative potential for small and medium-sized enterprises (SMEs), particularly in enhancing financial decision-making processes. However, SMEs often face significant barriers to implementing AI technologies, including limited resources, technical expertise, and data management capabilities. This paper presents a conceptual model for the adoption of AI in financial decision-making for SMEs. The proposed model addresses key challenges faced by SMEs, including limited resources, technical expertise, and data management capabilities. The model is structured into layers: data sources, data processing and integration, AI model deployment, decision support and automation, and validation and risk management. By implementing AI incrementally, SMEs can optimize financial forecasting, budgeting, investment strategies, and risk management. This paper highlights the importance of data quality and continuous model validation, providing a practical roadmap for SMEs to integrate AI into their financial operations. The study concludes with implications for SMEs adopting AI-driven financial processes and suggests areas for future research in AI applications for SME finance.

---

### One Detector Fits All: Robust and Adaptive Detection of Malicious Packages from PyPI to Enterprises

Authors: Biagio Montaruli, Luca Compagna, Serena Elisa Ponta, Davide Balzarotti
Published: 2025-12-03
Categories: cs.CR, cs.LG
arXiv: https://arxiv.org/abs/2512.04338v1

The rise of supply chain attacks via malicious Python packages demands robust detection solutions. Current approaches, however, overlook two critical challenges: robustness against adversarial source code transformations and adaptability to the varying false positive rate (FPR) requirements of different actors, from repository maintainers (requiring low FPR) to enterprise security teams (higher FPR tolerance). We introduce a robust detector capable of seamless integration into both public repositories like PyPI and enterprise ecosystems. To ensure robustness, we propose a novel methodology for generating adversarial packages using fine-grained code obfuscation. Combining these with adversarial training (AT) enhances detector robustness by 2.5x. We comprehensively evaluate AT effectiveness by testing our detector against 122,398 packages collected daily from PyPI over 80 days, showing that AT needs careful application: it makes the detector more robust to obfuscations and allows finding 10% more obfuscated packages, but slightly decreases performance on non-obfuscated packages. We demonstrate production adaptability of our detector via two case studies: (i) one for PyPI maintainers (tuned at 0.1% FPR) and (ii) one for enterprise teams (tuned at 10% FPR). In the former, we analyze 91,949 packages collected from PyPI over 37 days, achieving a daily detection rate of 2.48 malicious packages with only 2.18 false positives. In the latter, we analyze 1,596 packages adopted by a multinational software company, obtaining only 1.24 false positives daily. These results show that our detector can be seamlessly integrated into both public repositories like PyPI and enterprise ecosystems, ensuring a very low time budget of a few minutes to review the false positives. Overall, we uncovered 346 malicious packages, now reported to the community.

---

### RGE-GCN: Recursive Gene Elimination with Graph Convolutional Networks for RNA-seq based Early Cancer Detection

Authors: Shreyas Shende, Varsha Narayanan, Vishal Fenn, Yiran Huang, Dincer Goksuluk, Gaurav Choudhary, Melih Agraz, Mengjia Xu
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.04333v2

Early detection of cancer plays a key role in improving survival rates, but identifying reliable biomarkers from RNA-seq data is still a major challenge. The data are high-dimensional, and conventional statistical methods often fail to capture the complex relationships between genes. In this study, we introduce RGE-GCN (Recursive Gene Elimination with Graph Convolutional Networks), a framework that combines feature selection and classification in a single pipeline. Our approach builds a graph from gene expression profiles, uses a Graph Convolutional Network to classify cancer versus normal samples, and applies Integrated Gradients to highlight the most informative genes. By recursively removing less relevant genes, the model converges to a compact set of biomarkers that are both interpretable and predictive. We evaluated RGE-GCN on synthetic data as well as real-world RNA-seq cohorts of lung, kidney, and cervical cancers. Across all datasets, the method consistently achieved higher accuracy and F1-scores than standard tools such as DESeq2, edgeR, and limma-voom. Importantly, the selected genes aligned with well-known cancer pathways including PI3K-AKT, MAPK, SUMOylation, and immune regulation. These results suggest that RGE-GCN shows promise as a generalizable approach for RNA-seq based early cancer detection and biomarker discovery (https://rce-gcn.streamlit.app/ ).

---

### Data-regularized Reinforcement Learning for Diffusion Models at Scale

Authors: Haotian Ye, Kaiwen Zheng, Jiashu Xu, Puheng Li, Huayu Chen, Jiaqi Han, Sheng Liu, Qinsheng Zhang, Hanzi Mao, Zekun Hao, Prithvijit Chattopadhyay, Dinghao Yang et al.
Published: 2025-12-03
Categories: cs.LG
arXiv: https://arxiv.org/abs/2512.04332v3

Aligning generative diffusion models with human preferences via reinforcement learning (RL) is critical yet challenging. Most existing algorithms are often vulnerable to reward hacking, such as quality degradation, over-stylization, or reduced diversity. Our analysis demonstrates that this can be attributed to the inherent limitations of their regularization, which provides unreliable penalties. We introduce Data-regularized Diffusion Reinforcement Learning (DDRL), a novel framework that uses the forward KL divergence to anchor the policy to an off-policy data distribution. Theoretically, DDRL enables robust, unbiased integration of RL with standard diffusion training. Empirically, this translates into a simple yet effective algorithm that combines reward maximization with diffusion loss minimization. With over a million GPU hours of experiments and ten thousand double-blind human evaluations, we demonstrate on high-resolution video generation tasks that DDRL significantly improves rewards while alleviating the reward hacking seen in baselines, achieving the highest human preference and establishing a robust and scalable paradigm for diffusion post-training.

---

### Open Set Face Forgery Detection via Dual-Level Evidence Collection

Authors: Zhongyi Cai, Bryce Gernon, Wentao Bao, Yifan Li, Matthew Wright, Yu Kong
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04331v2

The surge in face forgeries has increasingly undermined confidence in the authenticity of online content. As generation algorithms rapidly evolve, new fake categories will constantly emerge, severely challenging existing face forgery detection methods. Although face forgery detection has recently improved, current techniques remain largely confined to binary Real-vs-Fake classification or the recognition of known fake categories. Moreover, they fail to identify the emergence of entirely new forgery methods. In this work, we study the Open Set Face Forgery Detection (OSFFD) problem, which requires the detection model to identify novel fake categories. To enhance its real-world applicability, we reformulate the OSFFD problem and address it through uncertainty estimation. Specifically, we propose the Dual-Level Evidential face forgery Detection (DLED) approach, which estimates prediction uncertainty by extracting and integrating category-specific evidence on the spatial and frequency levels. Comprehensive experiments across diverse settings demonstrate that our proposed DLED approach achieves state-of-the-art performance. Notably, it surpasses various existing baseline models by a $20\%$ margin on average when identifying forgeries from novel fake categories. Concurrently, our DLED method yields competitive performance on the standard binary Real-versus-Fake face forgery detection task.

---

### A Retrieval-Augmented Generation Approach to Extracting Algorithmic Logic from Neural Networks

Authors: Waleed Khalid, Dmitry Ignatov, Radu Timofte
Published: 2025-12-03
Categories: cs.CV, cs.SE
arXiv: https://arxiv.org/abs/2512.04329v2

Reusing existing neural-network components is central to research efficiency, yet discovering, extracting, and validating such modules across thousands of open-source repositories remains difficult. We introduce NN-RAG, a retrieval-augmented generation system that converts large, heterogeneous PyTorch codebases into a searchable and executable library of validated neural modules. Unlike conventional code search or clone-detection tools, NN-RAG performs scope-aware dependency resolution, import-preserving reconstruction, and validator-gated promotion -- ensuring that every retrieved block is scope-closed, compilable, and runnable. Applied to 19 major repositories, the pipeline extracted 1,289 candidate blocks, validated 941 (73.0%), and demonstrated that over 80% are structurally unique. Through multi-level de-duplication (exact, lexical, structural), we find that NN-RAG contributes the overwhelming majority of unique architectures to the LEMUR dataset, supplying approximately 72% of all novel network structures. Beyond quantity, NN-RAG uniquely enables cross-repository migration of architectural patterns, automatically identifying reusable modules in one project and regenerating them, dependency-complete, in another context. To our knowledge, no other open-source system provides this capability at scale. The framework's neutral specifications further allow optional integration with language models for synthesis or dataset registration without redistributing third-party code. Overall, NN-RAG transforms fragmented vision code into a reproducible, provenance-tracked substrate for algorithmic discovery, offering a first open-source solution that both quantifies and expands the diversity of executable neural architectures across repositories.

---

### DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle

Authors: Fangyu Lei, Jinxiang Meng, Yiming Huang, Junjie Zhao, Yitong Zhang, Jianwen Luo, Xin Zou, Ruiyi Yang, Wenbo Shi, Yan Gao, Shizhu He, Zuo Wang et al.
Published: 2025-12-03
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2512.04324v1

Real-world enterprise data intelligence workflows encompass data engineering that turns raw sources into analytical-ready tables and data analysis that convert those tables into decision-oriented insights. We introduce DAComp, a benchmark of 210 tasks that mirrors these complex workflows. Data engineering (DE) tasks require repository-level engineering on industrial schemas, including designing and building multi-stage SQL pipelines from scratch and evolving existing systems under evolving requirements. Data analysis (DA) tasks pose open-ended business problems that demand strategic planning, exploratory analysis through iterative coding, interpretation of intermediate results, and the synthesis of actionable recommendations. Engineering tasks are scored through execution-based, multi-metric evaluation. Open-ended tasks are assessed by a reliable, experimentally validated LLM-judge, which is guided by hierarchical, meticulously crafted rubrics. Our experiments reveal that even state-of-the-art agents falter on DAComp. Performance on DE tasks is particularly low, with success rates under 20%, exposing a critical bottleneck in holistic pipeline orchestration, not merely code generation. Scores on DA tasks also average below 40%, highlighting profound deficiencies in open-ended reasoning and demonstrating that engineering and analysis are distinct capabilities. By clearly diagnosing these limitations, DAComp provides a rigorous and realistic testbed to drive the development of truly capable autonomous data agents for enterprise settings. Our data and code are available at https://da-comp.github.io

---

### Bayes-DIC Net: Estimating Digital Image Correlation Uncertainty with Bayesian Neural Networks

Authors: Biao Chen, Zhenhua Lei, Yahui Zhang, Tongzhi Niu
Published: 2025-12-03
Categories: cs.CV, cs.AI, cs.CE, cs.LG
arXiv: https://arxiv.org/abs/2512.04323v1

This paper introduces a novel method for generating high-quality Digital Image Correlation (DIC) dataset based on non-uniform B-spline surfaces. By randomly generating control point coordinates, we construct displacement fields that encompass a variety of realistic displacement scenarios, which are subsequently used to generate speckle pattern datasets. This approach enables the generation of a large-scale dataset that capture real-world displacement field situations, thereby enhancing the training and generalization capabilities of deep learning-based DIC algorithms. Additionally, we propose a novel network architecture, termed Bayes-DIC Net, which extracts information at multiple levels during the down-sampling phase and facilitates the aggregation of information across various levels through a single skip connection during the up-sampling phase. Bayes-DIC Net incorporates a series of lightweight convolutional blocks designed to expand the receptive field and capture rich contextual information while minimizing computational costs. Furthermore, by integrating appropriate dropout modules into Bayes-DIC Net and activating them during the network inference stage, Bayes-DIC Net is transformed into a Bayesian neural network. This transformation allows the network to provide not only predictive results but also confidence levels in these predictions when processing real unlabeled datasets. This feature significantly enhances the practicality and reliability of our network in real-world displacement field prediction tasks. Through these innovations, this paper offers new perspectives and methods for dataset generation and algorithm performance enhancement in the field of DIC.

---

### MANTRA: a Framework for Multi-stage Adaptive Noise TReAtment During Training

Authors: Zixiao Zhao, Fatemeh H. Fard, Jie JW Wu
Published: 2025-12-03
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2512.04319v1

The reliable application of deep learning models to software engineering tasks hinges on high-quality training data. Yet, large-scale repositories inevitably introduce noisy or mislabeled examples that degrade both accuracy and robustness. While Noise Label Learning (NLL) has been extensively studied in other fields, there are a few works that investigate NLL in Software Engineering (SE) and Large Language Models (LLMs) for SE tasks. In this work, we propose MANTRA, a Multi-stage Adaptive Noise TReAtment framework that embeds noise diagnosis and mitigation directly into the fine-tuning process of code-Pretrained Language Models (PTM) and code-LLMs. We first investigate the effect of noise at varying levels on convergence and loss trajectories of the models. Then we apply an adaptive dropout strategy guided by per-sample loss dynamics and Gaussian Mixture Model clustering to exclude persistently noisy points while preserving clean data. Applying to code summarization and commit intent classification, our experiments reveal that some LLMs are more sensitive to noise than others. However, with MANTRA, the performance of all models in both tasks is improved. MANTRA enables researchers and practitioners to reduce the impact of errors introduced by the dataset in training, saves time in data cleaning and processing, while maximizing the effect of fine-tuning.

---

### SyncTrack4D: Cross-Video Motion Alignment and Video Synchronization for Multi-Video 4D Gaussian Splatting

Authors: Yonghan Lee, Tsung-Wei Huang, Shiv Gehlot, Jaehoon Choi, Guan-Ming Su, Dinesh Manocha
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04315v1

Modeling dynamic 3D scenes is challenging due to their high-dimensional nature, which requires aggregating information from multiple views to reconstruct time-evolving 3D geometry and motion. We present a novel multi-video 4D Gaussian Splatting (4DGS) approach designed to handle real-world, unsynchronized video sets. Our approach, SyncTrack4D, directly leverages dense 4D track representation of dynamic scene parts as cues for simultaneous cross-video synchronization and 4DGS reconstruction. We first compute dense per-video 4D feature tracks and cross-video track correspondences by Fused Gromov-Wasserstein optimal transport approach. Next, we perform global frame-level temporal alignment to maximize overlapping motion of matched 4D tracks. Finally, we achieve sub-frame synchronization through our multi-video 4D Gaussian splatting built upon a motion-spline scaffold representation. The final output is a synchronized 4DGS representation with dense, explicit 3D trajectories, and temporal offsets for each video. We evaluate our approach on the Panoptic Studio and SyncNeRF Blender, demonstrating sub-frame synchronization accuracy with an average temporal error below 0.26 frames, and high-fidelity 4D reconstruction reaching 26.3 PSNR scores on the Panoptic Studio dataset. To the best of our knowledge, our work is the first general 4D Gaussian Splatting approach for unsynchronized video sets, without assuming the existence of predefined scene objects or prior models.

---

### DisentangleFormer: Spatial-Channel Decoupling for Multi-Channel Vision

Authors: Jiashu Liao, Pietro Liò, Marc de Kamps, Duygu Sarikaya
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04314v1

Vision Transformers face a fundamental limitation: standard self-attention jointly processes spatial and channel dimensions, leading to entangled representations that prevent independent modeling of structural and semantic dependencies. This problem is especially pronounced in hyperspectral imaging, from satellite hyperspectral remote sensing to infrared pathology imaging, where channels capture distinct biophysical or biochemical cues. We propose DisentangleFormer, an architecture that achieves robust multi-channel vision representation through principled spatial-channel decoupling. Motivated by information-theoretic principles of decorrelated representation learning, our parallel design enables independent modeling of structural and semantic cues while minimizing redundancy between spatial and channel streams. Our design integrates three core components: (1) Parallel Disentanglement: Independently processes spatial-token and channel-token streams, enabling decorrelated feature learning across spatial and spectral dimensions, (2) Squeezed Token Enhancer: An adaptive calibration module that dynamically fuses spatial and channel streams, and (3) Multi-Scale FFN: complementing global attention with multi-scale local context to capture fine-grained structural and semantic dependencies. Extensive experiments on hyperspectral benchmarks demonstrate that DisentangleFormer achieves state-of-the-art performance, consistently outperforming existing models on Indian Pine, Pavia University, and Houston, the large-scale BigEarthNet remote sensing dataset, as well as an infrared pathology dataset. Moreover, it retains competitive accuracy on ImageNet while reducing computational cost by 17.8% in FLOPs. The code will be made publicly available upon acceptance.

---

### Mind-to-Face: Neural-Driven Photorealistic Avatar Synthesis via EEG Decoding

Authors: Haolin Xiong, Tianwen Fu, Pratusha Bhuvana Prasad, Yunxuan Cai, Haiwei Chen, Wenbin Teng, Hanyuan Xiao, Yajie Zhao
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04313v2

Current expressive avatar systems rely heavily on visual cues, failing when faces are occluded or when emotions remain internal. We present Mind-to-Face, the first framework that decodes non-invasive electroencephalogram (EEG) signals directly into high-fidelity facial expressions. We build a dual-modality recording setup to obtain synchronized EEG and multi-view facial video during emotion-eliciting stimuli, enabling precise supervision for neural-to-visual learning. Our model uses a CNN-Transformer encoder to map EEG signals into dense 3D position maps, capable of sampling over 65k vertices, capturing fine-scale geometry and subtle emotional dynamics, and renders them through a modified 3D Gaussian Splatting pipeline for photorealistic, view-consistent results. Through extensive evaluation, we show that EEG alone can reliably predict dynamic, subject-specific facial expressions, including subtle emotional responses, demonstrating that neural signals contain far richer affective and geometric information than previously assumed. Mind-to-Face establishes a new paradigm for neural-driven avatars, enabling personalized, emotion-aware telepresence and cognitive interaction in immersive environments.

---

### Real-time Cricket Sorting By Sex

Authors: Juan Manuel Cantarero Angulo, Matthew Smith
Published: 2025-12-03
Categories: cs.CV, q-bio.QM
arXiv: https://arxiv.org/abs/2512.04311v1

The global demand for sustainable protein sources is driving increasing interest in edible insects, with Acheta domesticus (house cricket) identified as one of the most suitable species for industrial production. Current farming practices typically rear crickets in mixed-sex populations without automated sex sorting, despite potential benefits such as selective breeding, optimized reproduction ratios, and nutritional differentiation. This work presents a low-cost, real-time system for automated sex-based sorting of Acheta domesticus, combining computer vision and physical actuation. The device integrates a Raspberry Pi 5 with the official Raspberry AI Camera and a custom YOLOv8 nano object detection model, together with a servo-actuated sorting arm. The model reached a mean Average Precision at IoU 0.5 (mAP@0.5) of 0.977 during testing, and real-world experiments with groups of crickets achieved an overall sorting accuracy of 86.8%. These results demonstrate the feasibility of deploying lightweight deep learning models on resource-constrained devices for insect farming applications, offering a practical solution to improve efficiency and sustainability in cricket production.

---

### RNNs perform task computations by dynamically warping neural representations

Authors: Arthur Pellegrino, Angus Chadwick
Published: 2025-12-03
Categories: cs.LG, math.DG, math.DS, q-bio.NC
arXiv: https://arxiv.org/abs/2512.04310v2

Analysing how neural networks represent data features in their activations can help interpret how they perform tasks. Hence, a long line of work has focused on mathematically characterising the geometry of such "neural representations." In parallel, machine learning has seen a surge of interest in understanding how dynamical systems perform computations on time-varying input data. Yet, the link between computation-through-dynamics and representational geometry remains poorly understood. Here, we hypothesise that recurrent neural networks (RNNs) perform computations by dynamically warping their representations of task variables. To test this hypothesis, we develop a Riemannian geometric framework that enables the derivation of the manifold topology and geometry of a dynamical system from the manifold of its inputs. By characterising the time-varying geometry of RNNs, we show that dynamic warping is a fundamental feature of their computations.

---

### Text-Only Training for Image Captioning with Retrieval Augmentation and Modality Gap Correction

Authors: Rui Fonseca, Bruno Martins, Gil Rocha
Published: 2025-12-03
Categories: cs.CV, cs.CL
arXiv: https://arxiv.org/abs/2512.04309v1

Image captioning has drawn considerable attention from the natural language processing and computer vision fields. Aiming to reduce the reliance on curated data, several studies have explored image captioning without any humanly-annotated image-text pairs for training, although existing methods are still outperformed by fully supervised approaches. This paper proposes TOMCap, i.e., an improved text-only training method that performs captioning without the need for aligned image-caption pairs. The method is based on prompting a pre-trained language model decoder with information derived from a CLIP representation, after undergoing a process to reduce the modality gap. We specifically tested the combined use of retrieved examples of captions, and latent vector representations, to guide the generation process. Through extensive experiments, we show that TOMCap outperforms other training-free and text-only methods. We also analyze the impact of different choices regarding the configuration of the retrieval-augmentation and modality gap reduction components.

---

### Evaluating Long-Context Reasoning in LLM-Based WebAgents

Authors: Andy Chung, Yichi Zhang, Kaixiang Lin, Aditya Rawal, Qiaozi Gao, Joyce Chai
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.04307v1

As large language model (LLM)-based agents become increasingly integrated into daily digital interactions, their ability to reason across long interaction histories becomes crucial for providing personalized and contextually aware assistance. However, the performance of these agents in long context scenarios, particularly for action-taking WebAgents operating in realistic web environments, remains largely unexplored. This paper introduces a benchmark for evaluating long context reasoning capabilities of WebAgents through sequentially dependent subtasks that require retrieval and application of information from extended interaction histories. We develop a novel evaluation framework that simulates multi-session user interactions by injecting irrelevant task trajectories between dependent subtasks, creating contexts ranging from 25,000 to 150,000 tokens. Through extensive evaluation of four popular models, Claude-3.7, GPT-4.1, Llama 4, and o4-mini, we observe a dramatic performance degradation as context length increases, with success rates dropping from 40-50\% in baseline conditions to less than 10\% in long context scenarios. Our detailed error analysis reveals that agents primarily fail due to getting stuck in loops and losing track of original task objectives. We further propose an implicit RAG approach that provides modest improvements by generating task-relevant summaries, though fundamental limitations in long context reasoning persist. These findings highlight critical challenges for deploying WebAgents in realistic, long-term user interaction scenarios and provide insights for developing more robust agent architectures capable of maintaining coherent task execution across extended contexts.

---

### How (Mis)calibrated is your Federated CLIP and what to do about it?

Authors: Mainak Singha, Masih Aminbeidokhti, Paolo Casari, Gianni Franchi, Elisa Ricci, Subhankar Roy
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04305v3

Vision-language models (VLMs) such as CLIP are increasingly adapted across decentralized data silos, yet the reliability of their predictions under federated learning (FL) remains largely unexplored. In this work, we present a systematic study of calibration in federated CLIP under non-IID client distributions. Our experiments reveal that widely used prompt-tuning methods consistently degrade calibration, often yielding substantially higher calibration error despite competitive recognition performance, while explicit training-time calibration regularizers provide only limited improvements. Motivated by these findings, we identify the choice of fine-tuning parameterization as a critical factor governing calibration and conduct a controlled comparison between prompt tuning and five backbone fine-tuning (BFT) strategies: AdaptFormer, LayerNorm, LoRA, VeRA, and DoRA. We find that BFT methods generally offer a more favorable accuracy-calibration trade-off than prompt tuning, although their benefits are not universal. Through extensive analysis, we show that calibration behavior is closely linked to the geometry of federated updates, residual parameterization, and the resulting client and logit drift. Across in-distribution, domain-generalization, and base-to-new evaluation settings, our results establish fine-tuning parameterization as a central design choice for building accurate and reliable federated CLIP models. Codes are available at https://github.com/mainaksingha01/FL2oRA.

---

### Gamma-from-Mono: Road-Relative, Metric, Self-Supervised Monocular Geometry for Vehicular Applications

Authors: Gasser Elazab, Maximilian Jansen, Michael Unterreiner, Olaf Hellwich
Published: 2025-12-03
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2512.04303v1

Accurate perception of the vehicle's 3D surroundings, including fine-scale road geometry, such as bumps, slopes, and surface irregularities, is essential for safe and comfortable vehicle control. However, conventional monocular depth estimation often oversmooths these features, losing critical information for motion planning and stability. To address this, we introduce Gamma-from-Mono (GfM), a lightweight monocular geometry estimation method that resolves the projective ambiguity in single-camera reconstruction by decoupling global and local structure. GfM predicts a dominant road surface plane together with residual variations expressed by gamma, a dimensionless measure of vertical deviation from the plane, defined as the ratio of a point's height above it to its depth from the camera, and grounded in established planar parallax geometry. With only the camera's height above ground, this representation deterministically recovers metric depth via a closed form, avoiding full extrinsic calibration and naturally prioritizing near-road detail. Its physically interpretable formulation makes it well suited for self-supervised learning, eliminating the need for large annotated datasets. Evaluated on KITTI and the Road Surface Reconstruction Dataset (RSRD), GfM achieves state-of-the-art near-field accuracy in both depth and gamma estimation while maintaining competitive global depth performance. Our lightweight 8.88M-parameter model adapts robustly across diverse camera setups and, to our knowledge, is the first self-supervised monocular approach evaluated on RSRD.

---

### Towards better dense rewards in Reinforcement Learning Applications

Authors: Shuyuan Zhang
Published: 2025-12-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2512.04302v1

Finding meaningful and accurate dense rewards is a fundamental task in the field of reinforcement learning (RL) that enables agents to explore environments more efficiently. In traditional RL settings, agents learn optimal policies through interactions with an environment guided by reward signals. However, when these signals are sparse, delayed, or poorly aligned with the intended task objectives, agents often struggle to learn effectively. Dense reward functions, which provide informative feedback at every step or state transition, offer a potential solution by shaping agent behavior and accelerating learning. Despite their benefits, poorly crafted reward functions can lead to unintended behaviors, reward hacking, or inefficient exploration. This problem is particularly acute in complex or high-dimensional environments where handcrafted rewards are difficult to specify and validate. To address this, recent research has explored a variety of approaches, including inverse reinforcement learning, reward modeling from human preferences, and self-supervised learning of intrinsic rewards. While these methods offer promising directions, they often involve trade-offs between generality, scalability, and alignment with human intent. This proposal explores several approaches to dealing with these unsolved problems and enhancing the effectiveness and reliability of dense reward construction in different RL applications.

---

### When do spectral gradient updates help in deep learning?

Authors: Damek Davis, Dmitriy Drusvyatskiy
Published: 2025-12-03
Categories: cs.LG, math.OC, stat.ML
arXiv: https://arxiv.org/abs/2512.04299v2

Spectral gradient methods, such as the recently popularized Muon optimizer, are a promising alternative to standard Euclidean gradient descent for training deep neural networks and transformers, but it is still unclear in which regimes they are expected to perform better. We propose a simple layerwise condition that predicts when a spectral update yields a larger decrease in the loss than a Euclidean gradient step. This condition compares, for each parameter block, the squared nuclear-to-Frobenius ratio of the gradient to the stable rank of the incoming activations. To understand when this condition may be satisfied, we first prove that post-activation matrices have low stable rank at Gaussian initialization in random feature regression, feedforward networks, and transformer blocks. In spiked random feature models we then show that, after a short burn-in, the Euclidean gradient's nuclear-to-Frobenius ratio grows with the data dimension while the stable rank of the activations remains bounded, so the predicted advantage of spectral updates scales with dimension. We validate these predictions in synthetic regression experiments and in NanoGPT-scale language model training, where we find that intermediate activations have low-stable-rank throughout training and the corresponding gradients maintain large nuclear-to-Frobenius ratios. Together, these results identify conditions for spectral gradient methods, such as Muon, to be effective in training deep networks and transformers.

---

### GRASP: GRouped Activation Shared Parameterization for Parameter-Efficient Fine-Tuning and Robust Inference of Transformers

Authors: Malyaban Bal, Abhronil Sengupta
Published: 2025-12-03
Categories: cs.LG, cs.NE
arXiv: https://arxiv.org/abs/2512.04296v2

Parameter-efficient fine-tuning (PEFT) provides a scalable alternative to full-model adaptation by updating only a small subset of parameters in large pre-trained models. We introduce GRASP - GRouped Activation Shared Parameterization - a lightweight PEFT framework that partitions the D-dimensional token representations of selected layers into K << D groups and learns a shared scaling and shifting vector for each group. This grouped modulation reduces the number of trainable parameters significantly while preserving the ability of the model to learn task-specific features. Building on this formulation, we further propose StochGRASP, which learns Gaussian distributions as perturbations to the pre-trained weights rather than deterministic values. This probabilistic parameterization along with a noise-aware loss function formulation enables modelling hardware-level variability in programmed weights and significantly improves robustness under non-ideal inference conditions-an important requirement for deployment on edge-based emerging AI hardware. Across GLUE (RoBERTa-base & RoBERTa-large) and E2E NLG (GPT-2 Medium), GRASP matches or exceeds the performance of established PEFT methods while achieving an order of magnitude reduction in trainable parameters compared to LoRA and BitFit. Under varying levels of noise, StochGRASP consistently outperforms deterministic variants, demonstrating its suitability for energy-efficient and noise-prone hardware platforms.

---

### SQuARE: Structured Query & Adaptive Retrieval Engine For Tabular Formats

Authors: Chinmay Gondhalekar, Urjitkumar Patel, Fang-Chun Yeh
Published: 2025-12-03
Categories: cs.CL
arXiv: https://arxiv.org/abs/2512.04292v1

Accurate question answering over real spreadsheets remains difficult due to multirow headers, merged cells, and unit annotations that disrupt naive chunking, while rigid SQL views fail on files lacking consistent schemas. We present SQuARE, a hybrid retrieval framework with sheet-level, complexity-aware routing. It computes a continuous score based on header depth and merge density, then routes queries either through structure-preserving chunk retrieval or SQL over an automatically constructed relational representation. A lightweight agent supervises retrieval, refinement, or combination of results across both paths when confidence is low. This design maintains header hierarchies, time labels, and units, ensuring that returned values are faithful to the original cells and straightforward to verify. Evaluated on multi-header corporate balance sheets, a heavily merged World Bank workbook, and diverse public datasets, SQuARE consistently surpasses single-strategy baselines and ChatGPT-4o on both retrieval precision and end-to-end answer accuracy while keeping latency predictable. By decoupling retrieval from model choice, the system is compatible with emerging tabular foundation models and offers a practical bridge toward a more robust table understanding.

---

### Artificial Intelligence Applications in Horizon Scanning for Infectious Diseases

Authors: Ian Miles, Mayumi Wakimoto, Wagner Meira, Daniela Paula, Daylene Ticiane, Bruno Rosa, Jane Biddulph, Stelios Georgiou, Valdir Ermida
Published: 2025-12-03
Categories: cs.AI, q-bio.PE
arXiv: https://arxiv.org/abs/2512.04287v1

This review explores the integration of Artificial Intelligence into Horizon Scanning, focusing on identifying and responding to emerging threats and opportunities linked to Infectious Diseases. We examine how AI tools can enhance signal detection, data monitoring, scenario analysis, and decision support. We also address the risks associated with AI adoption and propose strategies for effective implementation and governance. The findings contribute to the growing body of Foresight literature by demonstrating the potential and limitations of AI in Public Health preparedness.

---

### Learning Single-Image Super-Resolution in the JPEG Compressed Domain

Authors: Sruthi Srinivasan, Elham Shakibapour, Rajy Rawther, Mehdi Saeedi
Published: 2025-12-03
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2512.04284v1

Deep learning models have grown increasingly complex, with input data sizes scaling accordingly. Despite substantial advances in specialized deep learning hardware, data loading continues to be a major bottleneck that limits training and inference speed. To address this challenge, we propose training models directly on encoded JPEG features, reducing the computational overhead associated with full JPEG decoding and significantly improving data loading efficiency. While prior works have focused on recognition tasks, we investigate the effectiveness of this approach for the restoration task of single-image super-resolution (SISR). We present a lightweight super-resolution pipeline that operates on JPEG discrete cosine transform (DCT) coefficients in the frequency domain. Our pipeline achieves a 2.6x speedup in data loading and a 2.5x speedup in training, while preserving visual quality comparable to standard SISR approaches.

---

### Plug-and-Play Image Restoration with Flow Matching: A Continuous Viewpoint

Authors: Fan Jia, Yuhao Huang, Shih-Hsin Wang, Cristina Garcia-Cardona, Andrea L. Bertozzi, Bao Wang
Published: 2025-12-03
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2512.04283v1

Flow matching-based generative models have been integrated into the plug-and-play image restoration framework, and the resulting plug-and-play flow matching (PnP-Flow) model has achieved some remarkable empirical success for image restoration. However, the theoretical understanding of PnP-Flow lags its empirical success. In this paper, we derive a continuous limit for PnP-Flow, resulting in a stochastic differential equation (SDE) surrogate model of PnP-Flow. The SDE model provides two particular insights to improve PnP-Flow for image restoration: (1) It enables us to quantify the error for image restoration, informing us to improve step scheduling and regularize the Lipschitz constant of the neural network-parameterized vector field for error reduction. (2) It informs us to accelerate off-the-shelf PnP-Flow models via extrapolation, resulting in a rescaled version of the proposed SDE model. We validate the efficacy of the SDE-informed improved PnP-Flow using several benchmark tasks, including image denoising, deblurring, super-resolution, and inpainting. Numerical results show that our method significantly outperforms the baseline PnP-Flow and other state-of-the-art approaches, achieving superior performance across evaluation metrics.

---

### Inference-time Stochastic Refinement of GRU-Normalizing Flow for Real-time Video Motion Transfer

Authors: Tasmiah Haque, Srinjoy Das
Published: 2025-12-03
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2512.04282v2

Real-time video motion transfer applications such as immersive gaming and vision-based anomaly detection require accurate yet diverse future predictions to support realistic synthesis and robust downstream decision making under uncertainty. To improve the diversity of such sequential forecasts we propose a novel inference-time refinement technique that combines Gated Recurrent Unit-Normalizing Flows (GRU-NF) with stochastic sampling methods. While GRU-NF can capture multimodal distributions through its integration of normalizing flows within a temporal forecasting framework, its deterministic transformation structure can limit expressivity. To address this, inspired by Stochastic Normalizing Flows (SNF), we introduce Markov Chain Monte Carlo (MCMC) steps during GRU-NF inference, enabling the model to explore a richer output space and better approximate the true data distribution without retraining. We validate our approach in a keypoint-based video motion transfer pipeline, where capturing temporally coherent and perceptually diverse future trajectories is essential for realistic samples and low bandwidth communication. Experiments show that our inference framework, Gated Recurrent Unit- Stochastic Normalizing Flows (GRU-SNF) outperforms GRU-NF in generating diverse outputs without sacrificing accuracy, even under longer prediction horizons. By injecting stochasticity during inference, our approach captures multimodal behavior more effectively. These results highlight the potential of integrating stochastic dynamics with flow-based sequence models for generative time series forecasting. The code is available at: https://github.com/Tasmiah1408028/Inference-Time-Stochastic-Refinement-Of-GRU-NF-For-Real-Time-Video-Motion-Transfer

---

### Bootstrapped Mixed Rewards for RL Post-Training: Injecting Canonical Action Order

Authors: Prakhar Gupta, Vaibhav Gupta
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.04277v3

Post-training with reinforcement learning (RL) typically optimizes a single scalar objective and ignores structure in how solutions are produced. We ask whether a scalar hint toward a canonical solver ordering, used only during RL post-training, improves performance even when fine-tuned on randomized solution sequences. On Zebra puzzles, we fine-tune a Transformer on randomized solution orders, then post-train it with Group Relative Policy Optimization (GRPO) using two rewards: a sparse task reward that is 1 only when the puzzle is fully solved, and an ordering reward that increases when the model's emission order aligns with the canonical solver order. To compare signals cleanly, we combine them via fixed mixtures and use a simple bootstrapped scaling to equalize component magnitudes at initialization. Mixed rewards generally outperform task-only optimization, suggesting that coarse ordering signals can steer RL post-training toward canonical trajectories without modifying supervised data or architecture.

---

### The Geometry of Benchmarks: A New Path Toward AGI

Authors: Przemyslaw Chojecki
Published: 2025-12-03
Categories: cs.AI, cs.LG, math.ST
arXiv: https://arxiv.org/abs/2512.04276v1

Benchmarks are the primary tool for assessing progress in artificial intelligence (AI), yet current practice evaluates models on isolated test suites and provides little guidance for reasoning about generality or autonomous self-improvement. Here we introduce a geometric framework in which all psychometric batteries for AI agents are treated as points in a structured moduli space, and agent performance is described by capability functionals over this space. First, we define an Autonomous AI (AAI) Scale, a Kardashev-style hierarchy of autonomy grounded in measurable performance on batteries spanning families of tasks (for example reasoning, planning, tool use and long-horizon control). Second, we construct a moduli space of batteries, identifying equivalence classes of benchmarks that are indistinguishable at the level of agent orderings and capability inferences. This geometry yields determinacy results: dense families of batteries suffice to certify performance on entire regions of task space. Third, we introduce a general Generator-Verifier-Updater (GVU) operator that subsumes reinforcement learning, self-play, debate and verifier-based fine-tuning as special cases, and we define a self-improvement coefficient $κ$ as the Lie derivative of a capability functional along the induced flow. A variance inequality on the combined noise of generation and verification provides sufficient conditions for $κ> 0$. Our results suggest that progress toward artificial general intelligence (AGI) is best understood as a flow on moduli of benchmarks, driven by GVU dynamics rather than by scores on individual leaderboards.

---

### Quantitative Analysis of Technical Debt and Pattern Violation in Large Language Model Architectures

Authors: Tyler Slater
Published: 2025-12-03
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2512.04273v1

As Large Language Models (LLMs) transition from code completion tools to autonomous system architects, their impact on long-term software maintainability remains unquantified. While existing research benchmarks functional correctness (pass@k), this study presents the first empirical framework to measure "Architectural Erosion" and the accumulation of Technical Debt in AI-synthesized microservices. We conducted a comparative pilot study of three state-of-the-art models (GPT-5.1, Claude 4.5 Sonnet, and Llama 3 8B) by prompting them to implement a standardized Book Lending Microservice under strict Hexagonal Architecture constraints. Utilizing Abstract Syntax Tree (AST) parsing, we find that while proprietary models achieve high architectural conformance (0% violation rate for GPT-5.1), open-weights models exhibit critical divergence. Specifically, Llama 3 demonstrated an 80% Architectural Violation Rate, frequently bypassing interface adapters to create illegal circular dependencies between Domain and Infrastructure layers. Furthermore, we identified a phenomenon of "Implementation Laziness," where open-weights models generated 60% fewer Logical Lines of Code (LLOC) than their proprietary counterparts, effectively omitting complex business logic to satisfy token constraints. These findings suggest that without automated architectural linting, utilizing smaller open-weights models for system scaffolding accelerates the accumulation of structural technical debt.

---

### The Initialization Determines Whether In-Context Learning Is Gradient Descent

Authors: Shifeng Xie, Rui Yuan, Simone Rossi, Thomas Hannagan
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.04268v1

In-context learning (ICL) in large language models (LLMs) is a striking phenomenon, yet its underlying mechanisms remain only partially understood. Previous work connects linear self-attention (LSA) to gradient descent (GD), this connection has primarily been established under simplified conditions with zero-mean Gaussian priors and zero initialization for GD. However, subsequent studies have challenged this simplified view by highlighting its overly restrictive assumptions, demonstrating instead that under conditions such as multi-layer or nonlinear attention, self-attention performs optimization-like inference, akin to but distinct from GD. We investigate how multi-head LSA approximates GD under more realistic conditions specifically when incorporating non-zero Gaussian prior means in linear regression formulations of ICL. We first extend multi-head LSA embedding matrix by introducing an initial estimation of the query, referred to as the initial guess. We prove an upper bound on the number of heads needed for ICL linear regression setup. Our experiments confirm this result and further observe that a performance gap between one-step GD and multi-head LSA persists. To address this gap, we introduce yq-LSA, a simple generalization of single-head LSA with a trainable initial guess yq. We theoretically establish the capabilities of yq-LSA and provide experimental validation on linear regression tasks, thereby extending the theory that bridges ICL and GD. Finally, inspired by our findings in the case of linear regression, we consider widespread LLMs augmented with initial guess capabilities, and show that their performance is improved on a semantic similarity task.

---

### UniLight: A Unified Representation for Lighting

Authors: Zitian Zhang, Iliyan Georgiev, Michael Fischer, Yannick Hold-Geoffroy, Jean-François Lalonde, Valentin Deschaintre
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04267v2

Lighting has a strong influence on visual appearance, yet understanding and representing lighting in images remains notoriously difficult. Various lighting representations exist, such as environment maps, irradiance, spherical harmonics, or text, but they are incompatible, which limits cross-modal transfer. We thus propose UniLight, a joint latent space as lighting representation, that unifies multiple modalities within a shared embedding. Modality-specific encoders for text, images, irradiance, and environment maps are trained contrastively to align their representations, with an auxiliary spherical-harmonics prediction task reinforcing directional understanding. Our multi-modal data pipeline enables large-scale training and evaluation across three tasks: lighting-based retrieval, environment-map generation, and lighting control in diffusion-based image synthesis. Experiments show that our representation captures consistent and transferable lighting features, enabling flexible manipulation across modalities.

---

### Studying Various Activation Functions and Non-IID Data for Machine Learning Model Robustness

Authors: Long Dang, Thushari Hapuarachchi, Kaiqi Xiong, Jing Lin
Published: 2025-12-03
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2512.04264v1

Adversarial training is an effective method to improve the machine learning (ML) model robustness. Most existing studies typically consider the Rectified linear unit (ReLU) activation function and centralized training environments. In this paper, we study the ML model robustness using ten different activation functions through adversarial training in centralized environments and explore the ML model robustness in federal learning environments. In the centralized environment, we first propose an advanced adversarial training approach to improving the ML model robustness by incorporating model architecture change, soft labeling, simplified data augmentation, and varying learning rates. Then, we conduct extensive experiments on ten well-known activation functions in addition to ReLU to better understand how they impact the ML model robustness. Furthermore, we extend the proposed adversarial training approach to the federal learning environment, where both independent and identically distributed (IID) and non-IID data settings are considered. Our proposed centralized adversarial training approach achieves a natural and robust accuracy of 77.08% and 67.96%, respectively on CIFAR-10 against the fast gradient sign attacks. Experiments on ten activation functions reveal ReLU usually performs best. In the federated learning environment, however, the robust accuracy decreases significantly, especially on non-IID data. To address the significant performance drop in the non-IID data case, we introduce data sharing and achieve the natural and robust accuracy of 70.09% and 54.79%, respectively, surpassing the CalFAT algorithm, when 40% data sharing is used. That is, a proper percentage of data sharing can significantly improve the ML model robustness, which is useful to some real-world applications.

---

### Polynomiogram: An Integrated Framework for Root Visualization and Generative Art

Authors: Hoang Duc Nguyen, Anh Van Pham, Hien D. Nguyen
Published: 2025-12-03
Categories: cs.SE, cs.LG
arXiv: https://arxiv.org/abs/2512.04263v2

This work presents the Polynomiogram framework, an integrated computational platform for exploring, visualizing, and generating art from polynomial root systems. The main innovation is a flexible sampling scheme in which two independent parameters are drawn from user defined domains and mapped to the polynomial coefficients through a generating function. This design allows the same mathematical foundation to support both scientific investigation and generative algorithmic art. The framework integrates two complementary numerical engines: NumPy companion matrix solver for fast, large scale computation and MPSolve for high precision, scientifically rigorous validation. This dual architecture enables efficient visualization for creative use and accurate computation for research and education. Numerical accuracy was verified using classical ensembles, including the Kac and Lucas polynomials. The method was applied to the cubic polynomial system to analyze its bifurcation structure, demonstrating its value as both a scientific tool for exploring root phenomena and an educational aid for visualizing fundamental concepts in algebra and dynamical systems. Beyond analysis, the Polynomiogram also demonstrated its potential as a tool for personalized generative art. Examples include the use of the platform to generate a natural form resembling a hibiscus flower and to create personalized artwork expressing gratitude toward advances in artificial intelligence and large language models through a tribute composition.

---

### Catching UX Flaws in Code: Leveraging LLMs to Identify Usability Flaws at the Development Stage

Authors: Nolan Platt, Ethan Luchs, Sehrish Nizamani
Published: 2025-12-03
Categories: cs.SE, cs.AI, cs.HC
arXiv: https://arxiv.org/abs/2512.04262v1

Usability evaluations are essential for ensuring that modern interfaces meet user needs, yet traditional heuristic evaluations by human experts can be time-consuming and subjective, especially early in development. This paper investigates whether large language models (LLMs) can provide reliable and consistent heuristic assessments at the development stage. By applying Jakob Nielsen's ten usability heuristics to thirty open-source websites, we generated over 850 heuristic evaluations in three independent evaluations per site using a pipeline of OpenAI's GPT-4o. For issue detection, the model demonstrated moderate consistency, with an average pairwise Cohen's Kappa of 0.50 and an exact agreement of 84%. Severity judgments showed more variability: weighted Cohen's Kappa averaged 0.63, but exact agreement was just 56%, and Krippendorff's Alpha was near zero. These results suggest that while GPT-4o can produce internally consistent evaluations, especially for identifying the presence of usability issues, its ability to judge severity varies and requires human oversight in practice. Our findings highlight the feasibility and limitations of using LLMs for early-stage, automated usability testing, and offer a foundation for improving consistency in automated User Experience (UX) evaluation. To the best of our knowledge, our work provides one of the first quantitative inter-rater reliability analyses of automated heuristic evaluation and highlights methods for improving model consistency.

---

### Computational Linguistics Meets Libyan Dialect: A Study on Dialect Identification

Authors: Mansour Essgaer, Khamis Massud, Rabia Al Mamlook, Najah Ghmaid
Published: 2025-12-03
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2512.04257v1

This study investigates logistic regression, linear support vector machine, multinomial Naive Bayes, and Bernoulli Naive Bayes for classifying Libyan dialect utterances gathered from Twitter. The dataset used is the QADI corpus, which consists of 540,000 sentences across 18 Arabic dialects. Preprocessing challenges include handling inconsistent orthographic variations and non-standard spellings typical of the Libyan dialect. The chi-square analysis revealed that certain features, such as email mentions and emotion indicators, were not significantly associated with dialect classification and were thus excluded from further analysis. Two main experiments were conducted: (1) evaluating the significance of meta-features extracted from the corpus using the chi-square test and (2) assessing classifier performance using different word and character n-gram representations. The classification experiments showed that Multinomial Naive Bayes (MNB) achieved the highest accuracy of 85.89% and an F1-score of 0.85741 when using a (1,2) word n-gram and (1,5) character n-gram representation. In contrast, Logistic Regression and Linear SVM exhibited slightly lower performance, with maximum accuracies of 84.41% and 84.73%, respectively. Additional evaluation metrics, including log loss, Cohen kappa, and Matthew correlation coefficient, further supported the effectiveness of MNB in this task. The results indicate that carefully selected n-gram representations and classification models play a crucial role in improving the accuracy of Libyan dialect identification. This study provides empirical benchmarks and insights for future research in Arabic dialect NLP applications.

---

### Hey GPT-OSS, Looks Like You Got It -- Now Walk Me Through It! An Assessment of the Reasoning Language Models Chain of Thought Mechanism for Digital Forensics

Authors: Gaëtan Michelet, Janine Schneider, Aruna Withanage, Frank Breitinger
Published: 2025-12-03
Categories: cs.CR, cs.AI
arXiv: https://arxiv.org/abs/2512.04254v1

The use of large language models in digital forensics has been widely explored. Beyond identifying potential applications, research has also focused on optimizing model performance for forensic tasks through fine-tuning. However, limited result explainability reduces their operational and legal usability. Recently, a new class of reasoning language models has emerged, designed to handle logic-based tasks through an `internal reasoning' mechanism. Yet, users typically see only the final answer, not the underlying reasoning. One of these reasoning models is gpt-oss, which can be deployed locally, providing full access to its underlying reasoning process. This article presents the first investigation into the potential of reasoning language models for digital forensics. Four test use cases are examined to assess the usability of the reasoning component in supporting result explainability. The evaluation combines a new quantitative metric with qualitative analysis. Findings show that the reasoning component aids in explaining and validating language model outputs in digital forensics at medium reasoning levels, but this support is often limited, and higher reasoning levels do not enhance response quality.

---

### Fine-Tuning ChemBERTa for Predicting Inhibitory Activity Against TDP1 Using Deep Learning

Authors: Baichuan Zeng
Published: 2025-12-03
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2512.04252v1

Predicting the inhibitory potency of small molecules against Tyrosyl-DNA Phosphodiesterase 1 (TDP1)-a key target in overcoming cancer chemoresistance-remains a critical challenge in early drug discovery. We present a deep learning framework for the quantitative regression of pIC50 values from molecular Simplified Molecular Input Line Entry System (SMILES) strings using fine-tuned variants of ChemBERTa, a pre-trained chemical language model. Leveraging a large-scale consensus dataset of 177,092 compounds, we systematically evaluate two pre-training strategies-Masked Language Modeling (MLM) and Masked Token Regression (MTR)-under stratified data splits and sample weighting to address severe activity imbalance which only 2.1% are active. Our approach outperforms classical baselines Random Predictor in both regression accuracy and virtual screening utility, and has competitive performance compared to Random Forest, achieving high enrichment factor EF@1% 17.4 and precision Precision@1% 37.4 among top-ranked predictions. The resulting model, validated through rigorous ablation and hyperparameter studies, provides a robust, ready-to-deploy tool for prioritizing TDP1 inhibitors for experimental testing. By enabling accurate, 3D-structure-free pIC50 prediction directly from SMILES, this work demonstrates the transformative potential of chemical transformers in accelerating target-specific drug discovery.

---

### MVRoom: Controllable 3D Indoor Scene Generation with Multi-View Diffusion Models

Authors: Shaoheng Fang, Chaohui Yu, Fan Wang, Qixing Huang
Published: 2025-12-03
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2512.04248v1

We introduce MVRoom, a controllable novel view synthesis (NVS) pipeline for 3D indoor scenes that uses multi-view diffusion conditioned on a coarse 3D layout. MVRoom employs a two-stage design in which the 3D layout is used throughout to enforce multi-view consistency. The first stage employs novel representations to effectively bridge the 3D layout and consistent image-based condition signals for multi-view generation. The second stage performs image-conditioned multi-view generation, incorporating a layout-aware epipolar attention mechanism to enhance multi-view consistency during the diffusion process. Additionally, we introduce an iterative framework that generates 3D scenes with varying numbers of objects and scene complexities by recursively performing multi-view generation (MVRoom), supporting text-to-scene generation. Experimental results demonstrate that our approach achieves high-fidelity and controllable 3D scene generation for NVS, outperforming state-of-the-art baseline methods both quantitatively and qualitatively. Ablation studies further validate the effectiveness of key components within our generation pipeline.

---

### Toward Virtuous Reinforcement Learning: A Critique and Roadmap

Authors: Majid Ghasemi, Mark Crowley
Published: 2025-12-03
Categories: cs.AI
arXiv: https://arxiv.org/abs/2512.04246v2

This paper critiques common patterns in machine ethics for Reinforcement Learning (RL) and argues for a virtue focused alternative. We highlight two recurring limitations in much of the current literature: (i) rule based (deontological) methods that encode duties as constraints or shields often struggle under ambiguity and nonstationarity and do not cultivate lasting habits, and (ii) many reward based approaches, especially single objective RL, implicitly compress diverse moral considerations into a single scalar signal, which can obscure trade offs and invite proxy gaming in practice. We instead treat ethics as policy level dispositions, that is, relatively stable habits that hold up when incentives, partners, or contexts change. This shifts evaluation beyond rule checks or scalar returns toward trait summaries, durability under interventions, and explicit reporting of moral trade offs. Our roadmap combines four components: (1) social learning in multi agent RL to acquire virtue like patterns from imperfect but normatively informed exemplars; (2) multi objective and constrained formulations that preserve value conflicts and incorporate risk aware criteria to guard against harm; (3) affinity based regularization toward updateable virtue priors that support trait like stability under distribution shift while allowing norms to evolve; and (4) operationalizing diverse ethical traditions as practical control signals, making explicit the value and cultural assumptions that shape ethical RL benchmarks.

---

### 6 Fingers, 1 Kidney: Natural Adversarial Medical Images Reveal Critical Weaknesses of Vision-Language Models

Authors: Leon Mayer, Piotr Kalinowski, Caroline Ebersbach, Marcel Knopp, Tim Rädsch, Evangelia Christodoulou, Annika Reinke, Fiona R. Kolbinger, Lena Maier-Hein
Published: 2025-12-03
Categories: cs.CV
arXiv: https://arxiv.org/abs/2512.04238v2

Vision-language models (VLMs) are increasingly integrated into clinical workflows. However, existing benchmarks primarily assess performance on common anatomical presentations and fail to capture the challenges posed by rare variants. To address this gap, we introduce AdversarialAnatomyBench, the first benchmark comprising naturally occurring rare anatomical variants across diverse imaging modalities and anatomical regions. We call such variants that violate learned priors about "typical" human anatomy natural adversarial anatomy. Benchmarking 25 state-of-the-art VLMs with AdversarialAnatomyBench yielded three key insights. First, when queried with basic medical perception tasks, mean accuracy dropped from 71% on typical to 28% on atypical anatomy. Even the best-performing models, GPT-5, Gemini 2.5 Pro, and Llama 4 Maverick, showed performance drops of 41-51%. Second, model errors closely mirrored expected anatomical biases. Third, neither model scaling nor interventions, including bias-aware prompting and test-time reasoning, resolved these issues. These findings highlight a critical limitation in current VLMs: their poor generalization to rare anatomical presentations. AdversarialAnatomyBench provides a foundation for systematically measuring and mitigating anatomical bias in multimodal medical artificial intelligence (AI) systems.

---