# Research Papers — 2026-01 · bundle page 51 of 189 > Full-text markdown of papers #2501–2550 of 9403 published in 2026-01, newest first. AI-readable mirrors of the full 9403-paper 2026-01 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-01/page/51 [← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/50.md) [Next bundle →](https://aipod.dev/papers/archive/2026-01/page/52.md) ### From Cold Start to Active Learning: Embedding-Based Scan Selection for Medical Image Segmentation Authors: Devon Levy, Bar Assayag, Laura Gaspar, Ilan Shimshoni, Bella Specktor-Fadida Published: 2026-01-26 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2601.18532v2 Accurate segmentation annotations are critical for disease monitoring, yet manual labeling remains a major bottleneck due to the time and expertise required. Active learning (AL) alleviates this burden by prioritizing informative samples for annotation, typically through a diversity-based cold-start phase followed by uncertainty-driven selection. We propose a novel cold-start sampling strategy that combines foundation-model embeddings with clustering, including automatic selection of the number of clusters and proportional sampling across clusters, to construct a diverse and representative initial training. This is followed by an uncertainty-based AL framework that integrates spatial diversity to guide sample selection. The proposed method is intuitive and interpretable, enabling visualization of the feature-space distribution of candidate samples. We evaluate our approach on three datasets spanning X-ray and MRI modalities. On the CheXmask dataset, the cold-start strategy outperforms random selection, improving Dice from 0.918 to 0.929 and reducing the Hausdorff distance from 32.41 to 27.66 mm. In the AL setting, combined entropy and diversity selection improves Dice from 0.919 to 0.939 and reduces the Hausdorff distance from 30.10 to 19.16 mm. On the Montgomery dataset, cold-start gains are substantial, with Dice improving from 0.928 to 0.950 and Hausdorff distance decreasing from 14.22 to 9.38 mm. On the SynthStrip dataset, cold-start selection slightly affects Dice but reduces the Hausdorff distance from 9.43 to 8.69 mm, while active learning improves Dice from 0.816 to 0.826 and reduces the Hausdorff distance from 7.76 to 6.38 mm. Overall, the proposed framework consistently outperforms baseline methods in low-data regimes, improving segmentation accuracy. --- ### Exploring Fine-Tuning for In-Context Retrieval and Efficient KV-Caching in Long-Context Language Models Authors: Francesco Maria Molfese, Momchil Hardalov, Rexhina Blloshmi, Bill Byrne, Adrià de Gispert Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18527v1 With context windows of millions of tokens, Long-Context Language Models (LCLMs) can encode entire document collections, offering a strong alternative to conventional retrieval-augmented generation (RAG). However, it remains unclear whether fine-tuning strategies can improve long-context performance and translate to greater robustness under KV-cache compression techniques. In this work, we investigate which training strategies most effectively enhance LCLMs' ability to identify and use relevant information, as well as enhancing their robustness under KV-cache compression. Our experiments show substantial in-domain improvements, achieving gains of up to +20 points over the base model. However, out-of-domain generalization remains task dependent with large variance -- LCLMs excels on finance questions (+9 points), while RAG shows stronger performance on multiple-choice questions (+6 points) over the baseline models. Finally, we show that our fine-tuning approaches bring moderate improvements in robustness under KV-cache compression, with gains varying across tasks. --- ### Closing the Modality Gap Aligns Group-Wise Semantics Authors: Eleonora Grassucci, Giordano Cicchetti, Emanuele Frasca, Aurelio Uncini, Danilo Comminiello Published: 2026-01-26 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2601.18525v1 In multimodal learning, CLIP has been recognized as the \textit{de facto} method for learning a shared latent space across multiple modalities, placing similar representations close to each other and moving them away from dissimilar ones. Although CLIP-based losses effectively align modalities at the semantic level, the resulting latent spaces often remain only partially shared, revealing a structural mismatch known as the modality gap. While the necessity of addressing this phenomenon remains debated, particularly given its limited impact on instance-wise tasks (e.g., retrieval), we prove that its influence is instead strongly pronounced in group-level tasks (e.g., clustering). To support this claim, we introduce a novel method designed to consistently reduce this discrepancy in two-modal settings, with a straightforward extension to the general $n$-modal case. Through our extensive evaluation, we demonstrate our novel insight: while reducing the gap provides only marginal or inconsistent improvements in traditional instance-wise tasks, it significantly enhances group-wise tasks. These findings may reshape our understanding of the modality gap, highlighting its key role in improving performance on tasks requiring semantic grouping. --- ### From Human Labels to Literature: Semi-Supervised Learning of NMR Chemical Shifts at Scale Authors: Yongqi Jin, Yecheng Wang, Jun-jie Wang, Rong Zhu, Guolin Ke, Weinan E Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18524v2 Accurate prediction of nuclear magnetic resonance (NMR) chemical shifts is fundamental to spectral analysis and molecular structure elucidation, yet existing machine learning methods rely on limited, labor-intensive atom-assigned datasets. We propose a semi-supervised framework that learns NMR chemical shifts from millions of literature-extracted spectra without explicit atom-level assignments, integrating a small amount of labeled data with large-scale unassigned spectra. We formulate chemical shift prediction from literature spectra as a permutation-invariant set supervision problem, and show that under commonly satisfied conditions on the loss function, optimal bipartite matching reduces to a sorting-based loss, enabling stable large-scale semi-supervised training beyond traditional curated datasets. Our models achieve substantially improved accuracy and robustness over state-of-the-art methods and exhibit stronger generalization on significantly larger and more diverse molecular datasets. Moreover, by incorporating solvent information at scale, our approach captures systematic solvent effects across common NMR solvents for the first time. Overall, our results demonstrate that large-scale unlabeled spectra mined from the literature can serve as a practical and effective data source for training NMR shift models, suggesting a broader role of literature-derived, weakly structured data in data-centric AI for science. --- ### Scalable Transit Delay Prediction at City Scale: A Systematic Approach with Multi-Resolution Feature Engineering and Deep Learning Authors: Emna Boudabbous, Mohamed Karaa, Lokman Sboui, Julio Montecinos, Omar Alam Published: 2026-01-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2601.18521v1 Urban bus transit agencies need reliable, network-wide delay predictions to provide accurate arrival information to passengers and support real-time operational control. Accurate predictions help passengers plan their trips, reduce waiting time, and allow operations staff to adjust headways, dispatch extra vehicles, and manage disruptions. Although real-time feeds such as GTFS-Realtime (GTFS-RT) are now widely available, most existing delay prediction systems handle only a few routes, depend on hand-crafted features, and offer little guidance on how to design a scalable, reusable architecture. We present a city-scale prediction pipeline that combines multi-resolution feature engineering, dimensionality reduction, and deep learning. The framework generates 1,683 spatiotemporal features by exploring 23 aggregation combinations over H3 cells, routes, segments, and temporal patterns, and compresses them into 83 components using Adaptive PCA while preserving 95% of the variance. To avoid the "giant cluster" problem that occurs when dense urban areas fall into a single H3 region, we introduce a hybrid H3+topology clustering method that yields 12 balanced route clusters (coefficient of variation 0.608) and enables efficient distributed training. We compare five model architectures on six months of bus operations from the Société de transport de Montréal (STM) network in Montréal. A global LSTM with cluster-aware features achieves the best trade-off between accuracy and efficiency, outperforming transformer models by 18 to 52% while using 275 times fewer parameters. We also report multi-level evaluation at the elementary segment, segment, and trip level with walk-forward validation and latency analysis, showing that the proposed pipeline is suitable for real-time, city-scale deployment and can be reused for other networks with limited adaptation. --- ### GenAI for Social Work Field Education: Client Simulation with Real-Time Feedback Authors: James Sungarda, Hongkai Liu, Zilong Zhou, Tien-Hsuan Wu, Johnson Chun-Sing Cheung, Ben Kao Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18517v1 Field education is the signature pedagogy of social work, yet providing timely and objective feedback during training is constrained by the availability of instructors and counseling clients. In this paper, we present SWITCH, the Social Work Interactive Training Chatbot. SWITCH integrates realistic client simulation, real-time counseling skill classification, and a Motivational Interviewing (MI) progression system into the training workflow. To model a client, SWITCH uses a cognitively grounded profile comprising static fields (e.g., background, beliefs) and dynamic fields (e.g., emotions, automatic thoughts, openness), allowing the agent's behavior to evolve throughout a session realistically. The skill classification module identifies the counseling skills from the user utterances, and feeds the result to the MI controller that regulates the MI stage transitions. To enhance classification accuracy, we study in-context learning with retrieval over annotated transcripts, and a fine-tuned BERT multi-label classifier. In the experiments, we demonstrated that both BERT-based approach and in-context learning outperforms the baseline with big margin. SWITCH thereby offers a scalable, low-cost, and consistent training workflow that complements field education, and allows supervisors to focus on higher-level mentorship. --- ### LipNeXt: Scaling up Lipschitz-based Certified Robustness to Billion-parameter Models Authors: Kai Hu, Haoqi Hu, Matt Fredrikson Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18513v1 Lipschitz-based certification offers efficient, deterministic robustness guarantees but has struggled to scale in model size, training efficiency, and ImageNet performance. We introduce \emph{LipNeXt}, the first \emph{constraint-free} and \emph{convolution-free} 1-Lipschitz architecture for certified robustness. LipNeXt is built using two techniques: (1) a manifold optimization procedure that updates parameters directly on the orthogonal manifold and (2) a \emph{Spatial Shift Module} to model spatial pattern without convolutions. The full network uses orthogonal projections, spatial shifts, a simple 1-Lipschitz $β$-Abs nonlinearity, and $L_2$ spatial pooling to maintain tight Lipschitz control while enabling expressive feature mixing. Across CIFAR-10/100 and Tiny-ImageNet, LipNeXt achieves state-of-the-art clean and certified robust accuracy (CRA), and on ImageNet it scales to 1-2B large models, improving CRA over prior Lipschitz models (e.g., up to $+8\%$ at $\varepsilon{=}1$) while retaining efficient, stable low-precision training. These results demonstrate that Lipschitz-based certification can benefit from modern scaling trends without sacrificing determinism or efficiency. --- ### Using Large Language Models to Construct Virtual Top Managers: A Method for Organizational Research Authors: Antonio Garzon-Vico, Krithika Sharon Komalapati, Arsalan Shahid, Jan Rosier Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18512v1 This study introduces a methodological framework that uses large language models to create virtual personas of real top managers. Drawing on real CEO communications and Moral Foundations Theory, we construct LLM-based participants that simulate the decision-making of individual leaders. Across three phases, we assess construct validity, reliability, and behavioral fidelity by benchmarking these virtual CEOs against human participants. Our results indicate that theoretically scaffolded personas approximate the moral judgements observed in human samples, suggesting that LLM-based personas can serve as credible and complementary tools for organizational research in contexts where direct access to executives is limited. We conclude by outlining implications for future research using LLM-based personas in organizational settings. --- ### Just-In-Time Reinforcement Learning: Continual Learning in LLM Agents Without Gradient Updates Authors: Yibo Li, Zijie Lin, Ailin Deng, Xuan Zhang, Yufei He, Shuo Ji, Tri Cao, Bryan Hooi Published: 2026-01-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2601.18510v4 While Large Language Model (LLM) agents excel at general tasks, they inherently struggle with continual adaptation due to the frozen weights after deployment. Conventional reinforcement learning (RL) offers a solution but incurs prohibitive computational costs and the risk of catastrophic forgetting. We introduce Just-In-Time Reinforcement Learning (JitRL), a training-free framework that enables test-time policy optimization without any gradient updates. JitRL maintains a dynamic, non-parametric memory of experiences and retrieves relevant trajectories to estimate action advantages on-the-fly. These estimates are then used to directly modulate the LLM's output logits. We theoretically prove that this additive update rule is the exact closed-form solution to the KL-constrained policy optimization objective. Extensive experiments on WebArena and Jericho demonstrate that JitRL establishes a new state-of-the-art among training-free methods. Crucially, JitRL outperforms the performance of computationally expensive fine-tuning methods (e.g., WebRL) while reducing monetary costs by over 30 times, offering a scalable path for continual learning agents. The code is available at https://github.com/liushiliushi/JitRL. --- ### Conformal Prediction Algorithms for Time Series Forecasting: Methods and Benchmarking Authors: Andro Sabashvili Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18509v2 Reliable uncertainty quantification is of critical importance in time series forecasting, yet traditional methods often rely on restrictive distributional assumptions. Conformal prediction (CP) has emerged as a promising distribution-free framework for generating prediction intervals with rigorous theoretical guarantees. However, applying CP to sequential data presents a primary challenge: the temporal dependencies inherent in time series fundamentally violate the core assumption of data exchangeability, upon which standard CP guarantees are built. This paper critically examines the main categories of algorithmic solutions designed to address this conflict. We survey and benchmark methods that relax the exchangeability assumption, those that redefine the data unit to be a collection of independent time series, approaches that explicitly model the dynamics of the prediction residuals, and online learning algorithms that adapt to distribution shifts to maintain long-run coverage. We use AutoARIMA as the base forecaster on a large-scale monthly sales dataset, evaluating marginal coverage, interval width, and the Winkler score. Our benchmark results show that multi-step split conformal prediction method meets the 90% coverage threshold and demonstrates the best efficiency. --- ### Nearly Optimal Bayesian Inference for Structural Missingness Authors: Chen Liang, Donghua Yang, Yutong Zhao, Tianle Zhang, Shenghang Zhou, Zhiyu Liang, Hengtong Zhang, Hongzhi Wang, Ziqi Li, Xiyang Zhang, Zheng Liang, Yifei Li Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18500v2 Structural missingness breaks 'just impute and train': values can be undefined by causal or logical constraints, and the mask may depend on observed variables, unobserved variables (MNAR), and other missingness indicators. It simultaneously brings (i) a catch-22 situation with causal loop, prediction needs the missing features, yet inferring them depends on the missingness mechanism, (ii) under MNAR, the unseen are different, the missing part can come from a shifted distribution, and (iii) plug-in imputation, a single fill-in can lock in uncertainty and yield overconfident, biased decisions. In the Bayesian view, prediction via the posterior predictive distribution integrates over the full model posterior uncertainty, rather than relying on a single point estimate. This framework decouples (i) learning an in-model missing-value posterior from (ii) label prediction by optimizing the predictive posterior distribution, enabling posterior integration. This decoupling yields an in-model almost-free-lunch: once the posterior is learned, prediction is plug-and-play while preserving uncertainty propagation. It achieves SOTA on 43 classification and 15 imputation benchmarks, with finite-sample near Bayes-optimality guarantees under our SCM prior. --- ### DEEPMED: Building a Medical DeepResearch Agent via Multi-hop Med-Search Data and Turn-Controlled Agentic Training & Inference Authors: Zihan Wang, Hao Wang, Shi Feng, Xiaocui Yang, Daling Wang, Yiqun Zhang, Jinghao Lin, Haihua Yang, Xiaozhong Ji Published: 2026-01-26 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.18496v2 Medical reasoning models remain constrained by parametric knowledge and are thus susceptible to forgetting and hallucinations. DeepResearch (DR) models ground outputs in verifiable evidence from tools and perform strongly in general domains, but their direct transfer to medical field yields relatively limited gains. We attribute this to two gaps: task characteristic and tool-use scaling. Medical questions require evidence interpretation in a knowledge-intensive clinical context; while general DR models can retrieve information, they often lack clinical-context reasoning and thus "find it but fail to use it," leaving performance limited by medical abilities. Moreover, in medical scenarios, blindly scaling tool-call can inject noisy context, derailing sensitive medical reasoning and prompting repetitive evidence-seeking along incorrect paths. Therefore, we propose DeepMed. For data, we deploy a multi-hop med-search QA synthesis method supporting the model to apply the DR paradigm in medical contexts. For training, we introduce a difficulty-aware turn-penalty to suppress excessive tool-call growth. For inference, we bring a monitor to help validate hypotheses within a controlled number of steps and avoid context rot. Overall, on seven medical benchmarks, DeepMed improves its base model by 9.79\% on average and outperforms larger medical reasoning and DR models. --- ### DisasterInsight: A Building-Centric Benchmark for Evaluating Vision--Language Models in Disaster Response Authors: Sara Tehrani, Yonghao Xu, Leif Haglund, Amanda Berg, Gulnaz Zhambulova, Michael Felsberg Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18493v3 Vision--language models (VLMs) show promise for disaster-response remote sensing, but existing benchmarks mainly emphasize scene-level or damage-centric assessment. To study this building-centric gap, we introduce \method{}, a diagnostic benchmark built on xBD, a pre/post-disaster satellite dataset with building-level damage labels. \method{} enriches building instances with OpenStreetMap-derived functional labels and contains 134{,}108 task-specific instruction records across 15 task types, spanning instance-level assessment, scene-level counting, multi-instance reasoning, and structured report generation. The benchmark supports RGB pre/post-disaster imagery, single- and multi-view instance formulations, and scene-level RGB/SAR diagnostic inputs. Experiments with general-domain and remote-sensing VLMs show that models perform better on visible damage cues than on building-function understanding, multi-instance reasoning, counting, and grounded reporting. Instruction tuning improves performance on several tasks but does not close this building-centric gap. --- ### AgentDoG: A Diagnostic Guardrail Framework for AI Agent Safety and Security Authors: Dongrui Liu, Qihan Ren, Chen Qian, Shuai Shao, Yuejin Xie, Yu Li, Zhonghao Yang, Haoyu Luo, Peng Wang, Qingyu Liu, Binxin Hu, Ling Tang et al. Published: 2026-01-26 Categories: cs.AI, cs.CC, cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2601.18491v2 The rise of AI agents introduces complex safety and security challenges arising from autonomous tool use and environmental interactions. Current guardrail models lack agentic risk awareness and transparency in risk diagnosis. To introduce an agentic guardrail that covers complex and numerous risky behaviors, we first propose a unified three-dimensional taxonomy that orthogonally categorizes agentic risks by their source (where), failure mode (how), and consequence (what). Guided by this structured and hierarchical taxonomy, we introduce a new fine-grained agentic safety benchmark (ATBench) and a Diagnostic Guardrail framework for agent safety and security (AgentDoG). AgentDoG provides fine-grained and contextual monitoring across agent trajectories. More Crucially, AgentDoG can diagnose the root causes of unsafe actions and seemingly safe but unreasonable actions, offering provenance and transparency beyond binary labels to facilitate effective agent alignment. AgentDoG variants are available in three sizes (4B, 7B, and 8B parameters) across Qwen and Llama model families. Extensive experimental results demonstrate that AgentDoG achieves state-of-the-art performance in agentic safety moderation in diverse and complex interactive scenarios. All models and datasets are openly released. --- ### Different Demographic Cues Yield Inconsistent Conclusions About LLM Personalization and Bias Authors: Manuel Tonneau, Neil K. R. Sehgal, Niyati Malhotra, Sharif Kazemi, Victor Orozco-Olvera, Ana María Muñoz Boudet, Lakshmi Subramanian, Samuel P. Fraiberger, Sharath Chandra Guntuku, Valentin Hofmann Published: 2026-01-26 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2601.18486v3 Demographic cue-based evaluation is widely used to study how large language models (LLMs) adapt their responses to signaled demographic attributes within and across groups. This approach typically relies on a single cue (e.g., names) as a proxy for group membership, implicitly treating different cues as interchangeable operationalizations of a single underlying identity-conditioned behavior. We test this assumption in realistic advice-seeking interactions spanning 14.8 million prompts, focusing on race and gender in a U.S. context. We find that cues for the same group induce only partially overlapping changes in model responses, yielding inconsistent conclusions about personalization, while bias conclusions are unstable, with both magnitude and direction of group differences varying across cues. We further show that these inconsistencies reflect differences in cue-group association strength and linguistic features bundled within cues that shape model responses. Together, our findings suggest that demographic conditioning in LLMs is not a cue-invariant category-level parameter but depends fundamentally on how identity is cued, reflecting responses to linguistic signals rather than stable demographic categories. We therefore call for multi-cue, mechanism-aware evaluations as a foundation for robust and interpretable claims about demographic variation in LLM responses. --- ### Funny or Persuasive, but Not Both: Evaluating Fine-Grained Multi-Concept Control in LLMs Authors: Arya Labroo, Ivaxi Sheth, Vyas Raina, Amaani Ahmed, Mario Fritz Published: 2026-01-26 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2601.18483v1 Large Language Models (LLMs) offer strong generative capabilities, but many applications require explicit and \textit{fine-grained} control over specific textual concepts, such as humor, persuasiveness, or formality. Prior approaches in prompting and representation engineering can provide coarse or single-attribute control, but systematic evaluation of multi-attribute settings remains limited. We introduce an evaluation framework for fine-grained controllability for both single- and dual-concept scenarios, focusing on linguistically distinct concept pairs (e.g., persuasiveness vs.~humor). Surprisingly, across multiple LLMs and generative tasks, we find that performance often drops in the dual-concept setting, even though the chosen concepts should in principle be separable. This reveals a fundamental limitation of naive prompting-based control: models struggle with compositionality even when concepts are intuitively independent. Our framework provides systematic evidence of this gap and offers a principled approach for measuring the ability of future methods for multi-concept control. --- ### Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States Authors: Kyoleen Kwak, Hyoseok Hwang Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18479v1 Deep reinforcement learning has proven to be a powerful approach to solving control tasks, but its characteristic high-frequency oscillations make it difficult to apply in real-world environments. While prior methods have addressed action oscillations via architectural or loss-based methods, the latter typically depend on heuristic or synthetic definitions of state similarity to promote action consistency, which often fail to accurately reflect the underlying system dynamics. In this paper, we propose a novel loss-based method by introducing a transition-induced similar state. The transition-induced similar state is defined as the distribution of next states transitioned from the previous state. Since it utilizes only environmental feedback and actually collected data, it better captures system dynamics. Building upon this foundation, we introduce Action Smoothing by Aligning Actions with Predictions from Preceding States (ASAP), an action smoothing method that effectively mitigates action oscillations. ASAP enforces action smoothness by aligning the actions with those taken in transition-induced similar states and by penalizing second-order differences to suppress high-frequency oscillations. Experiments in Gymnasium and Isaac-Lab environments demonstrate that ASAP yields smoother control and improved policy performance over existing methods. --- ### LoD-Structured 3D Gaussian Splatting for Streaming Video Reconstruction Authors: Xinhui Liu, Can Wang, Lei Liu, Zhenghao Chen, Wei Jiang, Wei Wang, Dong Xu Published: 2026-01-26 Categories: cs.GR, cs.CV arXiv: https://arxiv.org/abs/2601.18475v1 Free-Viewpoint Video (FVV) reconstruction enables photorealistic and interactive 3D scene visualization; however, real-time streaming is often bottlenecked by sparse-view inputs, prohibitive training costs, and bandwidth constraints. While recent 3D Gaussian Splatting (3DGS) has advanced FVV due to its superior rendering speed, Streaming Free-Viewpoint Video (SFVV) introduces additional demands for rapid optimization, high-fidelity reconstruction under sparse constraints, and minimal storage footprints. To bridge this gap, we propose StreamLoD-GS, an LoD-based Gaussian Splatting framework designed specifically for SFVV. Our approach integrates three core innovations: 1) an Anchor- and Octree-based LoD-structured 3DGS with a hierarchical Gaussian dropout technique to ensure efficient and stable optimization while maintaining high-quality rendering; 2) a GMM-based motion partitioning mechanism that separates dynamic and static content, refining dynamic regions while preserving background stability; and 3) a quantized residual refinement framework that significantly reduces storage requirements without compromising visual fidelity. Extensive experiments demonstrate that StreamLoD-GS achieves competitive or state-of-the-art performance in terms of quality, efficiency, and storage. --- ### Latent Knowledge as a Predictor of Fact Acquisition in Fine-Tuned Large Language Models Authors: Daniel B. Hier, Tayo Obafemi-Ajayi Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18468v1 Large language models store biomedical facts with uneven strength after pretraining: some facts are present in the weights but are not reliably accessible under deterministic decoding (latent knowledge), while others are scarcely represented. We fine tuned Llama 3.1 8B Instruct to learn ontology term identifier mappings from the Human Phenotype Ontology (800 pairs) and the Gene Ontology (400 training pairs), withholding 400 GO pairs to test generalization. Treating learning as a time to event process across 20 epochs, we used stochastic decoding to detect latent knowledge at baseline and Cox proportional hazards models to identify predictors of acquisition, generalization, and degradation. Baseline deterministic recall for HPO was 2.8%, rising to 71.9% after fine-tuning. Latent knowledge was the strongest predictor of faster fact acquisition (HR 2.6) and was associated with earlier, higher peak learning rates and faster convergence; identifier frequency and curated annotation counts had smaller effects. Generalization to withheld GO facts was uncommon (5.8%) but more likely when latent knowledge was present. Previously correct GO mappings degraded more often for withheld (unseen) terms than for trained (seen) terms, suggesting a protective effect of reinforcement during training. These results show that latent knowledge predicts both the speed of factual learning during fine-tuning and the limited generalization of unseen ontology facts, while resistance to degradation depends on whether facts are reinforced. --- ### OffSeeker: Online Reinforcement Learning Is Not All You Need for Deep Research Agents Authors: Yuhang Zhou, Kai Zheng, Qiguang Chen, Mengkang Hu, Qingfeng Sun, Can Xu, Jingjing Chen Published: 2026-01-26 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2601.18467v2 Deep research agents have shown remarkable potential in handling long-horizon tasks. However, state-of-the-art performance typically relies on online reinforcement learning (RL), which is financially expensive due to extensive API calls. While offline training offers a more efficient alternative, its progress is hindered by the scarcity of high-quality research trajectories. In this paper, we demonstrate that expensive online reinforcement learning is not all you need to build powerful research agents. To bridge this gap, we introduce a fully open-source suite designed for effective offline training. Our core contributions include DeepForge, a ready-to-use task synthesis framework that generates large-scale research queries without heavy preprocessing; and a curated collection of 66k QA pairs, 33k SFT trajectories, and 21k DPO pairs. Leveraging these resources, we train OffSeeker (8B), a model developed entirely offline. Extensive evaluations across six benchmarks show that OffSeeker not only leads among similar-sized agents but also remains competitive with 30B-parameter systems trained via heavy online RL. --- ### Fair-Eye Net: A Fair, Trustworthy, Multimodal Integrated Glaucoma Full Chain AI System Authors: Wenbin Wei, Suyuan Yao, Cheng Huang, Xiangyu Gao Published: 2026-01-26 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2601.18464v1 Glaucoma is a top cause of irreversible blindness globally, making early detection and longitudinal follow-up pivotal to preventing permanent vision loss. Current screening and progression assessment, however, rely on single tests or loosely linked examinations, introducing subjectivity and fragmented care. Limited access to high-quality imaging tools and specialist expertise further compromises consistency and equity in real-world use. To address these gaps, we developed Fair-Eye Net, a fair, reliable multimodal AI system closing the clinical loop from glaucoma screening to follow-up and risk alerting. It integrates fundus photos, OCT structural metrics, VF functional indices, and demographic factors via a dual-stream heterogeneous fusion architecture, with an uncertainty-aware hierarchical gating strategy for selective prediction and safe referral. A fairness constraint reduces missed diagnoses in disadvantaged subgroups. Experimental results show it achieved an AUC of 0.912 (96.7% specificity), cut racial false-negativity disparity by 73.4% (12.31% to 3.28%), maintained stable cross-domain performance, and enabled 3-12 months of early risk alerts (92% sensitivity, 88% specificity). Unlike post hoc fairness adjustments, Fair-Eye Net optimizes fairness as a primary goal with clinical reliability via multitask learning, offering a reproducible path for clinical translation and large-scale deployment to advance global eye health equity. --- ### 3DGesPolicy: Phoneme-Aware Holistic Co-Speech Gesture Generation Based on Action Control Authors: Xuanmeng Sha, Liyun Zhang, Tomohiro Mashita, Naoya Chiba, Yuki Uranishi Published: 2026-01-26 Categories: cs.CV, cs.AI, cs.LG, cs.MM, cs.SD arXiv: https://arxiv.org/abs/2601.18451v1 Generating holistic co-speech gestures that integrate full-body motion with facial expressions suffers from semantically incoherent coordination on body motion and spatially unstable meaningless movements due to existing part-decomposed or frame-level regression methods, We introduce 3DGesPolicy, a novel action-based framework that reformulates holistic gesture generation as a continuous trajectory control problem through diffusion policy from robotics. By modeling frame-to-frame variations as unified holistic actions, our method effectively learns inter-frame holistic gesture motion patterns and ensures both spatially and semantically coherent movement trajectories that adhere to realistic motion manifolds. To further bridge the gap in expressive alignment, we propose a Gesture-Audio-Phoneme (GAP) fusion module that can deeply integrate and refine multi-modal signals, ensuring structured and fine-grained alignment between speech semantics, body motion, and facial expressions. Extensive quantitative and qualitative experiments on the BEAT2 dataset demonstrate the effectiveness of our 3DGesPolicy across other state-of-the-art methods in generating natural, expressive, and highly speech-aligned holistic gestures. --- ### On Procrustes Contamination in Machine Learning Applications of Geometric Morphometrics Authors: Lloyd Austin Courtenay Published: 2026-01-26 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2601.18448v1 Geometric morphometrics (GMM) is widely used to quantify shape variation, more recently serving as input for machine learning (ML) analyses. Standard practice aligns all specimens via Generalized Procrustes Analysis (GPA) prior to splitting data into training and test sets, potentially introducing statistical dependence and contaminating downstream predictive models. Here, the effects of GPA-induced contamination are formally characterised using controlled 2D and 3D simulations across varying sample sizes, landmark densities, and allometric patterns. A novel realignment procedure is proposed, whereby test specimens are aligned to the training set prior to model fitting, eliminating cross-sample dependency. Simulations reveal a robust "diagonal" in sample-size vs. landmark-space, reflecting the scaling of RMSE under isotropic variation, with slopes analytically derived from the degrees of freedom in Procrustes tangent space. The importance of spatial autocorrelation among landmarks is further demonstrated using linear and convolutional regression models, highlighting performance degradation when landmark relationships are ignored. This work establishes the need for careful preprocessing in ML applications of GMM, provides practical guidelines for realignment, and clarifies fundamental statistical constraints inherent to Procrustes shape space. --- ### GCFX: Generative Counterfactual Explanations for Deep Graph Models at the Model Level Authors: Jinlong Hu, Jiacheng Liu Published: 2026-01-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2601.18447v1 Deep graph learning models have demonstrated remarkable capabilities in processing graph-structured data and have been widely applied across various fields. However, their complex internal architectures and lack of transparency make it difficult to explain their decisions, resulting in opaque models that users find hard to understand and trust. In this paper, we explore model-level explanation techniques for deep graph learning models, aiming to provide users with a comprehensive understanding of the models' overall decision-making processes and underlying mechanisms. Specifically, we address the problem of counterfactual explanations for deep graph learning models by introducing a generative model-level counterfactual explanation approach called GCFX, which is based on deep graph generation. This approach generates a set of high-quality counterfactual explanations that reflect the model's global predictive behavior by leveraging an enhanced deep graph generation framework and a global summarization algorithm. GCFX features an architecture that combines dual encoders, structure-aware taggers, and Message Passing Neural Network decoders, enabling it to accurately learn the true latent distribution of input data and generate high-quality, closely related counterfactual examples. Subsequently, a global counterfactual summarization algorithm selects the most representative and comprehensive explanations from numerous candidate counterfactuals, providing broad insights into the model's global predictive patterns. Experiments on a synthetic dataset and several real-world datasets demonstrate that GCFX outperforms existing methods in terms of counterfactual validity and coverage while maintaining low explanation costs, thereby offering crucial support for enhancing the practicality and trustworthiness of global counterfactual explanations. --- ### Beyond Speedups: Hardware-Aware Evaluation of Evolutionary Algorithms on GPUs Authors: Xinmeng Yu, Tao Jiang, Ran Cheng, Yaochu Jin, Kay Chen Tan Published: 2026-01-26 Categories: cs.NE arXiv: https://arxiv.org/abs/2601.18446v3 Evolutionary algorithms (EAs) are increasingly executed on graphics processing units (GPUs) to exploit population-level parallelism. This shift changes the resource model under which EAs are designed and evaluated. However, many GPU-based EA studies still focus mainly on implementation-level speedup after porting CPU-oriented algorithms to GPUs, providing limited insight into how algorithmic mechanisms, function-evaluation (FE) budgets, population scales, and hardware utilization jointly affect optimization behavior. In response, this paper goes beyond speedup measurement and studies the scaling behavior of EAs on GPUs from a hardware-aware evaluation perspective. We evaluate 16 representative EAs on 30 benchmark problems across CPU and GPU platforms, covering single-objective optimization, multi-objective optimization, numerical benchmarks, and neuroevolution tasks. The study leads to four findings. First, GPU acceleration is highly heterogeneous across algorithms because different evolutionary mechanisms expose different degrees of batched computation, memory regularity, and synchronization. Second, FE-budgeted evaluation remains useful for measuring sample efficiency, but it provides only a limited observation window under GPU execution; time-budgeted evaluation is therefore necessary for assessing practical time-to-solution and long-horizon search behavior. Third, GPU effectiveness depends on scaling regimes induced by problem dimension and population size, where parallelism may be underutilized, effective, or saturated. Fourth, GPU execution makes very large populations practically affordable, and several evolutionary mechanisms can convert this increased population scale into improved optimization performance. These results indicate that GPU parallelism should not be treated only as a post hoc acceleration tool, but as part of the evaluation and design assumptions of scalable EAs. --- ### Fusion of Spatio-Temporal and Multi-Scale Frequency Features for Dry Electrodes MI-EEG Decoding Authors: Tianyi Gong, Can Han, Junxi Wu, Dahong Qian Published: 2026-01-26 Categories: cs.HC, cs.LG arXiv: https://arxiv.org/abs/2601.18424v1 Dry-electrode Motor Imagery Electroencephalography (MI-EEG) enables fast, comfortable, real-world Brain Computer Interface by eliminating gels and shortening setup for at-home and wearable use.However, dry recordings pose three main issues: lower Signal-to-Noise Ratio with more baseline drift and sudden transients; weaker and noisier data with poor phase alignment across trials; and bigger variances between sessions. These drawbacks lead to larger data distribution shift, making features less stable for MI-EEG tasks.To address these problems, we introduce STGMFM, a tri-branch framework tailored for dry-electrode MI-EEG, which models complementary spatio-temporal dependencies via dual graph orders, and captures robust envelope dynamics with a multi-scale frequency mixing branch, motivated by the observation that amplitude envelopes are less sensitive to contact variability than instantaneous waveforms. Physiologically meaningful connectivity priors guide learning, and decision-level fusion consolidates a noise-tolerant consensus. On our collected dry-electrode MI-EEG, STGMFM consistently surpasses competitive CNN/Transformer/graph baselines. Codes are available at https://github.com/Tianyi-325/STGMFM. --- ### Gradient Regularized Natural Gradients Authors: Satya Prakash Dash, Hossein Abdi, Wei Pan, Samuel Kaski, Mingfei Sun Published: 2026-01-26 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2601.18420v2 Gradient regularization (GR) has been shown to improve the generalizability of trained models. While Natural Gradient Descent has been shown to accelerate optimization in the initial phase of training, little attention has been paid to how the training dynamics of second-order optimizers can benefit from GR. In this work, we propose Gradient-Regularized Natural Gradients (GRNG), a family of scalable second-order optimizers that integrate explicit gradient regularization with natural gradient updates. Our framework introduces two frequentist algorithms: Regularized Explicit Natural Gradient (RENG), which utilizes double backpropagation to explicitly minimize the gradient norm, and Regularized Implicit Natural Gradient (RING), which incorporates regularization implicitly into the update direction. We also propose a Bayesian variant based on a Regularized-Kalman formulation that eliminates the need for FIM inversion entirely. We establish convergence guarantees for GRNG, showing that gradient regularization improves stability and enables convergence to global minima. Empirically, we demonstrate that GRNG consistently enhances both optimization speed and generalization compared to first-order methods (SGD, AdamW) and second-order baselines (K-FAC, Sophia), with strong results on vision and language benchmarks. --- ### Emergent Cooperation in Quantum Multi-Agent Reinforcement Learning Using Communication Authors: Michael Kölle, Christian Reff, Leo Sünkel, Julian Hager, Gerhard Stenzel, Claudia Linnhoff-Popien Published: 2026-01-26 Categories: quant-ph, cs.AI, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2601.18419v1 Emergent cooperation in classical Multi-Agent Reinforcement Learning has gained significant attention, particularly in the context of Sequential Social Dilemmas (SSDs). While classical reinforcement learning approaches have demonstrated capability for emergent cooperation, research on extending these methods to Quantum Multi-Agent Reinforcement Learning remains limited, particularly through communication. In this paper, we apply communication approaches to quantum Q-Learning agents: the Mutual Acknowledgment Token Exchange (MATE) protocol, its extension Mutually Endorsed Distributed Incentive Acknowledgment Token Exchange (MEDIATE), the peer rewarding mechanism Gifting, and Reinforced Inter-Agent Learning (RIAL). We evaluate these approaches in three SSDs: the Iterated Prisoner's Dilemma, Iterated Stag Hunt, and Iterated Game of Chicken. Our experimental results show that approaches using MATE with temporal-difference measure (MATE\textsubscript{TD}), AutoMATE, MEDIATE-I, and MEDIATE-S achieved high cooperation levels across all dilemmas, demonstrating that communication is a viable mechanism for fostering emergent cooperation in Quantum Multi-Agent Reinforcement Learning. --- ### daVinci-Dev: Agent-native Mid-training for Software Engineering Authors: Ji Zeng, Dayuan Fu, Tiantian Mi, Yumin Zhuang, Yaxing Huang, Xuefeng Li, Lyumanshan Ye, Muhang Xie, Qishuo Hua, Zhen Huang, Mohan Jiang, Hanning Wang et al. Published: 2026-01-26 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2601.18418v2 Recently, the frontier of Large Language Model (LLM) capabilities has shifted from single-turn code generation to agentic software engineering-a paradigm where models autonomously navigate, edit, and test complex repositories. While post-training methods have become the de facto approach for code agents, **agentic mid-training**-mid-training (MT) on large-scale data that mirrors authentic agentic workflows-remains critically underexplored due to substantial resource requirements, despite offering a more scalable path to instilling foundational agentic behaviors than relying solely on expensive reinforcement learning. A central challenge in realizing effective agentic mid-training is the distribution mismatch between static training data and the dynamic, feedback-rich environment of real development. To address this, we present a systematic study of agentic mid-training, establishing both the data synthesis principles and training methodology for effective agent development at scale. Central to our approach is **agent-native data**-supervision comprising two complementary types of trajectories: **contextually-native trajectories** that preserve the complete information flow an agent experiences, offering broad coverage and diversity; and **environmentally-native trajectories** collected from executable repositories where observations stem from actual tool invocations and test executions, providing depth and interaction authenticity. We verify the model's agentic capabilities on `SWE-Bench Verified`. We demonstrate our superiority over the previous open software engineering mid-training recipe `Kimi-Dev` under two post-training settings with an aligned base model and agentic scaffold, while using less than half mid-training tokens (73.1B). Besides relative advantage, our best performing 32B and 72B models achieve **56.1%** and **58.5%** resolution rates, respectively, which are ... --- ### Pisets: A Robust Speech Recognition System for Lectures and Interviews Authors: Ivan Bondarenko, Daniil Grebenkin, Oleg Sedukhin, Mikhail Klementev, Roman Derunets, Lyudmila Budneva Published: 2026-01-26 Categories: cs.CL, cs.SD, eess.AS arXiv: https://arxiv.org/abs/2601.18415v1 This work presents a speech-to-text system "Pisets" for scientists and journalists which is based on a three-component architecture aimed at improving speech recognition accuracy while minimizing errors and hallucinations associated with the Whisper model. The architecture comprises primary recognition using Wav2Vec2, false positive filtering via the Audio Spectrogram Transformer (AST), and final speech recognition through Whisper. The implementation of curriculum learning methods and the utilization of diverse Russian-language speech corpora significantly enhanced the system's effectiveness. Additionally, advanced uncertainty modeling techniques were introduced, contributing to further improvements in transcription quality. The proposed approaches ensure robust transcribing of long audio data across various acoustic conditions compared to WhisperX and the usual Whisper model. The source code of "Pisets" system is publicly available at GitHub: https://github.com/bond005/pisets. --- ### Comparative Evaluation of Machine Learning Algorithms for Affective State Recognition from Children's Drawings Authors: Aura Loredana Dan Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18414v1 Autism spectrum disorder (ASD) represents a neurodevelopmental condition characterized by difficulties in expressing emotions and communication, particularly during early childhood. Understanding the affective state of children at an early age remains challenging, as conventional assessment methods are often intrusive, subjective, or difficult to apply consistently. This paper builds upon previous work on affective state recognition from children's drawings by presenting a comparative evaluation of machine learning models for emotion classification. Three deep learning architectures -- MobileNet, EfficientNet, and VGG16 -- are evaluated within a unified experimental framework to analyze classification performance, robustness, and computational efficiency. The models are trained using transfer learning on a dataset of children's drawings annotated with emotional labels provided by psychological experts. The results highlight important trade-offs between lightweight and deeper architectures when applied to drawing-based affective computing tasks, particularly in mobile and real-time application contexts. --- ### Frequency-Based Hyperparameter Selection in Games Authors: Aniket Sanyal, Baraah A. M. Sidahmed, Rebekka Burkholz, Tatjana Chavdarova Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18409v1 Learning in smooth games fundamentally differs from standard minimization due to rotational dynamics, which invalidate classical hyperparameter tuning strategies. Despite their practical importance, effective methods for tuning in games remain underexplored. A notable example is LookAhead (LA), which achieves strong empirical performance but introduces additional parameters that critically influence performance. We propose a principled approach to hyperparameter selection in games by leveraging frequency estimation of oscillatory dynamics. Specifically, we analyze oscillations both in continuous-time trajectories and through the spectrum of the discrete dynamics in the associated frequency-based space. Building on this analysis, we introduce \emph{Modal LookAhead (MoLA)}, an extension of LA that selects the hyperparameters adaptively to a given problem. We provide convergence guarantees and demonstrate in experiments that MoLA accelerates training in both purely rotational games and mixed regimes, all with minimal computational overhead. --- ### Larger than memory image processing Authors: Jon Sporring, David Stansby Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18407v1 This report addresses larger-than-memory image analysis for petascale datasets such as 1.4 PB electron-microscopy volumes and 150 TB human-organ atlases. We argue that performance is fundamentally I/O-bound. We show that structuring analysis as streaming passes over data is crucial. For 3D volumes, two representations are popular: stacks of 2D slices (e.g., directories or multi-page TIFF) and 3D chunked layouts (e.g., Zarr/HDF5). While for a few algorithms, chunked layout on disk is crucial to keep disk I/O at a minimum, we show how the slice-based streaming architecture can be built on top of either image representation in a manner that minimizes disk I/O. This is in particular advantageous for algorithms relying on neighbouring values, since the slicing streaming architecture is 1D, which implies that there are only 2 possible sweeping orders, both of which are aligned with the order in which images are read from the disk. This is in contrast to 3D chunks, in which any sweep cannot be done without accessing each chunk at least 9 times. We formalize this with sweep-based execution (natural 2D/3D orders), windowed operations, and overlap-aware tiling to minimize redundant access. Building on these principles, we introduce a domain-specific language (DSL) that encodes algorithms with intrinsic knowledge of their optimal streaming and memory use; the DSL performs compile-time and run-time pipeline analyses to automatically select window sizes, fuse stages, tee and zip streams, and schedule passes for limited-RAM machines, yielding near-linear I/O scans and predictable memory footprints. The approach integrates with existing tooling for segmentation and morphology but reframes pre/post-processing as pipelines that privilege sequential read/write patterns, delivering substantial throughput gains for extremely large images without requiring full-volume residency in memory. --- ### Superlinear Multi-Step Attention Authors: Yufeng Huang Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18401v1 In this paper, we propose \textbf{Superlinear attention}, a fully trainable multi-step attention architecture that achieves subquadratic complexity for long sequences while preserving \textbf{random context access} (a.k.a.\ structural non-exclusion): no eligible token position is structurally excluded from being selected for attention. Superlinear attention reformulates standard causal self-attention as a multi-step search problem with $N$ steps, yielding an overall complexity of $O(L^{1+\frac{1}{N}})$. To illustrate the architecture, we present a baseline $N=2$ implementation, which is algorithmically analogous to standard jump search. In this $O(L^{3/2})$ instantiation, the first step performs $O(L^{3/2})$ span-search to select relevant spans of the sequence, and the second step applies $O(L^{3/2})$ span-attention (standard attention restricted to the selected spans). In an upscaled $O(L^{1.54})$ configuration for robustness, we achieve an average decoding throughput of 114 tokens/sec at 1M context length and 80 tokens/sec at 10M context in our implementation on a modified 30B hybrid MoE model on a single B200 GPU. With limited training, we also obtain strong performance on the NIAH (Needle In A Haystack) task up to 256K context length, demonstrating that the routed span selection is learnable end-to-end. This paper emphasizes architectural formulation, scaling analysis, and systems feasibility, and presents initial validation; comprehensive quality evaluations across diverse long-context tasks are left to future work. --- ### Estimating Dense-Packed Zone Height in Liquid-Liquid Separation: A Physics-Informed Neural Network Approach Authors: Mehmet Velioglu, Song Zhai, Alexander Mitsos, Adel Mhamdi, Andreas Jupke, Manuel Dahmen Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18399v2 Separating liquid-liquid dispersions in gravity settlers is critical in chemical, pharmaceutical, and recycling processes. The dense-packed zone height is an important performance and safety indicator but it is often expensive and impractical to measure due to optical limitations. We propose a framework to estimate phase heights by combining a PINN model with readily available volume flow measurements, without requiring phase height measurements during deployment. To this end, a physics-informed neural network (PINN) is first pretrained on synthetic data and physics equations derived from a low-fidelity (approximate) mechanistic model to reduce the need for extensive experimental data. While the mechanistic model is used to generate synthetic training data, only volume balance equations are used in the PINN, as incorporating droplet coalescence and sedimentation submodels would be computationally prohibitive. The pretrained PINN is then fine-tuned with scarce experimental phase height and flow-rate data to capture the actual dynamics of the separator. We then deploy the differentiable PINN as a predictive model in an Extended Kalman Filter inspired state estimation framework, enabling the phase heights to be tracked and updated using flow-rate measurements only. We first test the two-stage trained PINN by forward simulation from a known initial state against the mechanistic model and a non-pretrained PINN. We then evaluate phase height estimation performance with the filter, comparing the two-stage trained PINN with a two-stage trained purely data-driven neural network. All model types are trained and evaluated using ensembles to account for model parameter uncertainty. In all evaluations, the two-stage trained PINN yields the most accurate phase-height estimates. --- ### Noise-Robust AV-ASR Using Visual Features Both in the Whisper Encoder and Decoder Authors: Zhengyang Li, Thomas Graave, Björn Möller, Zehang Wu, Matthias Franz, Tim Fingscheidt Published: 2026-01-26 Categories: eess.AS, cs.CL, cs.CV, cs.SD arXiv: https://arxiv.org/abs/2601.18396v1 In audiovisual automatic speech recognition (AV-ASR) systems, information fusion of visual features in a pre-trained ASR has been proven as a promising method to improve noise robustness. In this work, based on the prominent Whisper ASR, first, we propose a simple and effective visual fusion method -- use of visual features both in encoder and decoder (dual-use) -- to learn the audiovisual interactions in the encoder and to weigh modalities in the decoder. Second, we compare visual fusion methods in Whisper models of various sizes. Our proposed dual-use method shows consistent noise robustness improvement, e.g., a 35% relative improvement (WER: 4.41% vs. 6.83%) based on Whisper small, and a 57% relative improvement (WER: 4.07% vs. 9.53%) based on Whisper medium, compared to typical reference middle fusion in babble noise with a signal-to-noise ratio (SNR) of 0dB. Third, we conduct ablation studies examining the impact of various module designs and fusion options. Fine-tuned on 1929 hours of audiovisual data, our dual-use method using Whisper medium achieves 4.08% (MUSAN babble noise) and 4.43% (NoiseX babble noise) average WER across various SNRs, thereby establishing a new state-of-the-art in noisy conditions on the LRS3 AV-ASR benchmark. Our code is at https://github.com/ifnspaml/Dual-Use-AVASR --- ### Do not be greedy, Think Twice: Sampling and Selection for Document-level Information Extraction Authors: Mikel Zubillaga, Oscar Sainz, Oier Lopez de Lacalle, Eneko Agirre Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18395v3 Document-level Information Extraction (DocIE) aims to produce an output template with the entities, relations, and events of interest occurring in the given document. Standard practices include prompting decoder-only LLMs using greedy decoding to avoid output variability. Rather than treating this variability as a limitation, we show that sampling can produce substantially better solutions than greedy decoding, especially when using reasoning models. We thus propose ThinkTwice, a sampling and selection framework in which the LLM generates multiple candidate templates for a given document, and a selection module chooses the most suitable one. We introduce both an unsupervised method that exploits agreement across generated outputs, and a supervised selection method using reward models trained on labeled DocIE data. To address the scarcity of golden reasoning trajectories for DocIE, we propose a rejection-sampling-based method to generate silver training data that pairs output templates with reasoning traces. Our experiments show the validity of unsupervised and supervised ThinkTwice, consistently outperforming greedy baselines and the supervised state-of-the-art. --- ### OCR-Enhanced Multimodal ASR Can Read While Listening Authors: Junli Chen, Changli Tang, Yixuan Li, Guangzhi Sun, Chao Zhang Published: 2026-01-26 Categories: cs.SD, cs.CL, eess.AS arXiv: https://arxiv.org/abs/2601.18393v1 Visual information, such as subtitles in a movie, often helps automatic speech recognition. In this paper, we propose Donut-Whisper, an audio-visual ASR model with dual encoder to leverage visual information to improve speech recognition performance in both English and Chinese. Donut-Whisper combines the advantage of the linear and the Q-Former-based modality alignment structures via a cross-attention module, generating more powerful audio-visual features. Meanwhile, we propose a lightweight knowledge distillation scheme showcasing the potential of using audio-visual models to teach audio-only models to achieve better performance. Moreover, we propose a new multilingual audio-visual speech recognition dataset based on movie clips containing both Chinese and English partitions. As a result, Donut-Whisper achieved significantly better performance on both English and Chinese partition of the dataset compared to both Donut and Whisper large V3 baselines. In particular, an absolute 5.75% WER reduction and a 16.5% absolute CER reduction were achieved on the English and Chinese sets respectively compared to the Whisper ASR baseline. --- ### Efficient Complex-Valued Vision Transformers for MRI Classification Directly from k-Space Authors: Moritz Rempe, Lukas T. Rotkopf, Marco Schlimbach, Helmut Becker, Fabian Hörst, Johannes Haubold, Philipp Dammann, Kevin Kröninger, Jens Kleesiek Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18392v1 Deep learning applications in Magnetic Resonance Imaging (MRI) predominantly operate on reconstructed magnitude images, a process that discards phase information and requires computationally expensive transforms. Standard neural network architectures rely on local operations (convolutions or grid-patches) that are ill-suited for the global, non-local nature of raw frequency-domain (k-Space) data. In this work, we propose a novel complex-valued Vision Transformer (kViT) designed to perform classification directly on k-Space data. To bridge the geometric disconnect between current architectures and MRI physics, we introduce a radial k-Space patching strategy that respects the spectral energy distribution of the frequency-domain. Extensive experiments on the fastMRI and in-house datasets demonstrate that our approach achieves classification performance competitive with state-of-the-art image-domain baselines (ResNet, EfficientNet, ViT). Crucially, kViT exhibits superior robustness to high acceleration factors and offers a paradigm shift in computational efficiency, reducing VRAM consumption during training by up to 68$\times$ compared to standard methods. This establishes a pathway for resource-efficient, direct-from-scanner AI analysis. --- ### ARMOR: Agentic Reasoning for Methods Orchestration and Reparameterization for Robust Adversarial Attacks Authors: Gabriel Lee Jun Rong, Christos Korgialas, Dion Jia Xu Ho, Pai Chet Ng, Xiaoxiao Miao, Konstantinos N. Plataniotis Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18386v1 Existing automated attack suites operate as static ensembles with fixed sequences, lacking strategic adaptation and semantic awareness. This paper introduces the Agentic Reasoning for Methods Orchestration and Reparameterization (ARMOR) framework to address these limitations. ARMOR orchestrates three canonical adversarial primitives, Carlini-Wagner (CW), Jacobian-based Saliency Map Attack (JSMA), and Spatially Transformed Attacks (STA) via Vision Language Models (VLM)-guided agents that collaboratively generate and synthesize perturbations through a shared ``Mixing Desk". Large Language Models (LLMs) adaptively tune and reparameterize parallel attack agents in a real-time, closed-loop system that exploits image-specific semantic vulnerabilities. On standard benchmarks, ARMOR achieves improved cross-architecture transfer and reliably fools both settings, delivering a blended output for blind targets and selecting the best attack or blended attacks for white-box targets using a confidence-and-SSIM score. --- ### Estimation of geometric transformation matrices using grid-shaped pilot signals Authors: Rinka Kawano, Masaki Kawamura Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18385v1 Digital watermarking techniques are essential to prevent unauthorized use of images. Since pirated images are often geometrically distorted by operations such as scaling and cropping, accurate synchronization - detecting the embedding position of the watermark - is critical for proper extraction. In particular, cropping changes the origin of the image, making synchronization difficult. However, few existing methods are robust against cropping. To address this issue, we propose a watermarking method that estimates geometric transformations applied to a stego image using a pilot signal, allowing synchronization even after cropping. A grid-shaped pilot signal with distinct horizontal and vertical values is embedded in the image. When the image is transformed, the grid is also distorted. By analyzing this distortion, the transformation matrix can be estimated. Applying the Radon transform to the distorted image allows estimation of the grid angles and intervals. In addition, since the horizontal and vertical grid lines are encoded differently, the grid orientation can be determined, which reduces ambiguity. To validate our method, we performed simulations with anisotropic scaling, rotation, shearing, and cropping. The results show that the proposed method accurately estimates transformation matrices with low error under both single and composite attacks. --- ### Dynamic Thinking-Token Selection for Efficient Reasoning in Large Reasoning Models Authors: Zhenyuan Guo, Tong Chen, Wenlong Meng, Chen Gong, Xin Yu, Chengkun Wei, Wenzhi Chen Published: 2026-01-26 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2601.18383v2 Large Reasoning Models (LRMs) excel at solving complex problems by explicitly generating a reasoning trace before deriving the final answer. However, these extended generations incur substantial memory footprint and computational overhead, bottlenecking LRMs' efficiency. This work uses attention maps to analyze the influence of reasoning traces and uncover an interesting phenomenon: only some decision-critical tokens in a reasoning trace steer the model toward the final answer, while the remaining tokens contribute negligibly. Building on this observation, we propose Dynamic Thinking-Token Selection (DynTS). This method identifies decision-critical tokens and retains only their associated Key-Value (KV) cache states during inference, evicting the remaining redundant entries to optimize efficiency. --- ### AI Agent for Reverse-Engineering Legacy Finite-Difference Code and Translating to Devito Authors: Yinghan Hou, Zongyou Yang Published: 2026-01-26 Categories: cs.AI, cs.SE arXiv: https://arxiv.org/abs/2601.18381v2 To facilitate the transformation of legacy finite difference implementations into the Devito environment, this study develops an integrated AI agent framework. Retrieval-Augmented Generation (RAG) and open-source Large Language Models are combined through multi-stage iterative workflows in the system's hybrid LangGraph architecture. The agent constructs an extensive Devito knowledge graph through document parsing, structure-aware segmentation, extraction of entity relationships, and Leiden-based community detection. GraphRAG optimisation enhances query performance across semantic communities that include seismic wave simulation, computational fluid dynamics, and performance tuning libraries. A reverse engineering component derives three-level query strategies for RAG retrieval through static analysis of Fortran source code. To deliver precise contextual information for language model guidance, the multi-stage retrieval pipeline performs parallel searching, concept expansion, community-scale retrieval, and semantic similarity analysis. Code synthesis is governed by Pydantic-based constraints to guarantee structured outputs and reliability. A comprehensive validation framework integrates conventional static analysis with the G-Eval approach, covering execution correctness, structural soundness, mathematical consistency, and API compliance. The overall agent workflow is implemented on the LangGraph framework and adopts concurrent processing to support quality-based iterative refinement and state-aware dynamic routing. The principal contribution lies in the incorporation of feedback mechanisms motivated by reinforcement learning, enabling a transition from static code translation toward dynamic and adaptive analytical behavior. --- ### Corpus-Based Approaches to Igbo Diacritic Restoration Authors: Ignatius Ezeani Published: 2026-01-26 Categories: cs.CL, cs.CY, cs.IR arXiv: https://arxiv.org/abs/2601.18380v1 With natural language processing (NLP), researchers aim to enable computers to identify and understand patterns in human languages. This is often difficult because a language embeds many dynamic and varied properties in its syntax, pragmatics and phonology, which need to be captured and processed. The capacity of computers to process natural languages is increasing because NLP researchers are pushing its boundaries. But these research works focus more on well-resourced languages such as English, Japanese, German, French, Russian, Mandarin Chinese, etc. Over 95% of the world's 7000 languages are low-resourced for NLP, i.e. they have little or no data, tools, and techniques for NLP work. In this thesis, we present an overview of diacritic ambiguity and a review of previous diacritic disambiguation approaches on other languages. Focusing on the Igbo language, we report the steps taken to develop a flexible framework for generating datasets for diacritic restoration. Three main approaches, the standard n-gram model, the classification models and the embedding models were proposed. The standard n-gram models use a sequence of previous words to the target stripped word as key predictors of the correct variants. For the classification models, a window of words on both sides of the target stripped word was used. The embedding models compare the similarity scores of the combined context word embeddings and the embeddings of each of the candidate variant vectors. --- ### Hierarchical Text Classification with LLM-Refined Taxonomies Authors: Jonas Golde, Nicolaas Jedema, Ravi Krishnan, Phong Le Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18375v1 Hierarchical text classification (HTC) depends on taxonomies that organize labels into structured hierarchies. However, many real-world taxonomies introduce ambiguities, such as identical leaf names under similar parent nodes, which prevent language models (LMs) from learning clear decision boundaries. In this paper, we present TaxMorph, a framework that uses large language models (LLMs) to transform entire taxonomies through operations such as renaming, merging, splitting, and reordering. Unlike prior work, our method revises the full hierarchy to better match the semantics encoded by LMs. Experiments across three HTC benchmarks show that LLM-refined taxonomies consistently outperform human-curated ones in various settings up to +2.9pp. in F1. To better understand these improvements, we compare how well LMs can assign leaf nodes to parent nodes and vice versa across human-curated and LLM-refined taxonomies. We find that human-curated taxonomies lead to more easily separable clusters in embedding space. However, the LLM-refined taxonomies align more closely with the model's actual confusion patterns during classification. In other words, even though they are harder to separate, they better reflect the model's inductive biases. These findings suggest that LLM-guided refinement creates taxonomies that are more compatible with how models learn, improving HTC performance. --- ### CitiLink: Enhancing Municipal Transparency and Citizen Engagement through Searchable Meeting Minutes Authors: Rodrigo Silva, José Evans, José Isidro, Miguel Marques, Afonso Fonseca, Ricardo Morais, João Canavilhas, Arian Pasquali, Purificação Silvano, Alípio Jorge, Nuno Guimarães, Sérgio Nunes et al. Published: 2026-01-26 Categories: cs.CL arXiv: https://arxiv.org/abs/2601.18374v3 City council minutes are typically lengthy and formal documents with a bureaucratic writing style. Although publicly available, their structure often makes it difficult for citizens or journalists to efficiently find information. In this demo, we present CitiLink, a platform designed to transform unstructured municipal meeting minutes into structured and searchable data, demonstrating how NLP and IR can enhance the accessibility and transparency of local government. The system employs LLMs to extract metadata, discussed subjects, and voting outcomes, which are then indexed in a database to support full-text search with BM25 ranking and faceted filtering through a user-friendly interface. The developed system was built over a collection of 120 minutes made available by six Portuguese municipalities. To assess its usability, CitiLink was tested through guided sessions with municipal personnel, providing insights into how real users interact with the system. In addition, we evaluated Gemini's performance in extracting relevant information from the minutes, highlighting its effectiveness in data extraction. --- ### Gaze Prediction in Virtual Reality Without Eye Tracking Using Visual and Head Motion Cues Authors: Christos Petrou, Harris Partaourides, Athanasios Balomenos, Yannis Kopsinis, Sotirios Chatzis Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18372v1 Gaze prediction plays a critical role in Virtual Reality (VR) applications by reducing sensor-induced latency and enabling computationally demanding techniques such as foveated rendering, which rely on anticipating user attention. However, direct eye tracking is often unavailable due to hardware limitations or privacy concerns. To address this, we present a novel gaze prediction framework that combines Head-Mounted Display (HMD) motion signals with visual saliency cues derived from video frames. Our method employs UniSal, a lightweight saliency encoder, to extract visual features, which are then fused with HMD motion data and processed through a time-series prediction module. We evaluate two lightweight architectures, TSMixer and LSTM, for forecasting future gaze directions. Experiments on the EHTask dataset, along with deployment on commercial VR hardware, show that our approach consistently outperforms baselines such as Center-of-HMD and Mean Gaze. These results demonstrate the effectiveness of predictive gaze modeling in reducing perceptual lag and enhancing natural interaction in VR environments where direct eye tracking is constrained. --- ### OREHAS: A fully automated deep-learning pipeline for volumetric endolymphatic hydrops quantification in MRI Authors: Caterina Fuster-Barceló, Claudia Castrillón, Laura Rodrigo-Muñoz, Victor Manuel Suárez-Vega, Nicolás Pérez-Fernández, Gorka Bastarrika, Arrate Muñoz-Barrutia Published: 2026-01-26 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.18368v2 We present OREHAS (Optimized Recognition & Evaluation of volumetric Hydrops in the Auditory System), the first fully automatic pipeline for volumetric quantification of endolymphatic hydrops (EH) from routine 3D-SPACE-MRC and 3D-REAL-IR MRI. The system integrates three components -- slice classification, inner ear localization, and sequence-specific segmentation -- into a single workflow that computes per-ear endolymphatic-to-vestibular volume ratios (ELR) directly from whole MRI volumes, eliminating the need for manual intervention. Trained with only 3 to 6 annotated slices per patient, OREHAS generalized effectively to full 3D volumes, achieving Dice scores of 0.90 for SPACE-MRC and 0.75 for REAL-IR. In an external validation cohort with complete manual annotations, OREHAS closely matched expert ground truth (VSI = 74.3%) and substantially outperformed the clinical syngo.via software (VSI = 42.5%), which tended to overestimate endolymphatic volumes. Across 19 test patients, vestibular measurements from OREHAS were consistent with syngo.via, while endolymphatic volumes were systematically smaller and more physiologically realistic. These results show that reliable and reproducible EH quantification can be achieved from standard MRI using limited supervision. By combining efficient deep-learning-based segmentation with a clinically aligned volumetric workflow, OREHAS reduces operator dependence, ensures methodological consistency. Besides, the results are compatible with established imaging protocols. The approach provides a robust foundation for large-scale studies and for recalibrating clinical diagnostic thresholds based on accurate volumetric measurements of the inner ear. --- ### Making medical vision-language models think causally across modalities with retrieval-augmented cross-modal reasoning Authors: Weiqin Yang, Haowen Xue, Qingyi Peng, Hexuan Hu, Qian Huang, Tingbo Zhang Published: 2026-01-26 Categories: cs.LG arXiv: https://arxiv.org/abs/2601.18356v1 Medical vision-language models (VLMs) achieve strong performance in diagnostic reporting and image-text alignment, yet their underlying reasoning mechanisms remain fundamentally correlational, exhibiting reliance on superficial statistical associations that fail to capture the causal pathophysiological mechanisms central to clinical decision-making. This limitation makes them fragile, prone to hallucinations, and sensitive to dataset biases. Retrieval-augmented generation (RAG) offers a partial remedy by grounding predictions in external knowledge. However, conventional RAG depends on semantic similarity, introducing new spurious correlations. We propose Multimodal Causal Retrieval-Augmented Generation, a framework that integrates causal inference principles with multimodal retrieval. It retrieves clinically relevant exemplars and causal graphs from external sources, conditioning model reasoning on counterfactual and interventional evidence rather than correlations alone. Applied to radiology report generation, diagnosis prediction, and visual question answering, it improves factual accuracy, robustness to distribution shifts, and interpretability. Our results highlight causal retrieval as a scalable path toward medical VLMs that think beyond pattern matching, enabling trustworthy multimodal reasoning in high-stakes clinical settings. --- ### Can Good Writing Be Generative? Expert-Level AI Writing Emerges through Fine-Tuning on High-Quality Books Authors: Tuhin Chakrabarty, Paramveer S. Dhillon Published: 2026-01-26 Categories: cs.AI, cs.CL, cs.HC arXiv: https://arxiv.org/abs/2601.18353v1 Creative writing has long been considered a uniquely human endeavor, requiring voice and style that machines could not replicate. This assumption is challenged by Generative AI that can emulate thousands of author styles in seconds with negligible marginal labor. To understand this better, we conducted a behavioral experiment where 28 MFA writers (experts) competed against three LLMs in emulating 50 critically acclaimed authors. Based on blind pairwise comparisons by 28 expert judges and 131 lay judges, we find that experts preferred human writing in 82.7% of cases under the in-context prompting condition but this reversed to 62% preference for AI after fine-tuning on authors' complete works. Lay judges, however, consistently preferred AI writing. Debrief interviews with expert writers revealed that their preference for AI writing triggered an identity crisis, eroding aesthetic confidence and questioning what constitutes "good writing." These findings challenge discourse about AI's creative limitations and raise fundamental questions about the future of creative labor. ---