# Research Papers — 2026-01 · bundle page 86 of 189 > Full-text markdown of papers #4251–4300 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/86 [← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/85.md) [Next bundle →](https://aipod.dev/papers/archive/2026-01/page/87.md) ### Preconditioning Benefits of Spectral Orthogonalization in Muon Authors: Jianhao Ma, Yu Huang, Yuejie Chi, Yuxin Chen Published: 2026-01-20 Categories: cs.LG, cs.AI, math.OC, stat.ML arXiv: https://arxiv.org/abs/2601.13474v1 The Muon optimizer, a matrix-structured algorithm that leverages spectral orthogonalization of gradients, is a milestone in the pretraining of large language models. However, the underlying mechanisms of Muon -- particularly the role of gradient orthogonalization -- remain poorly understood, with very few works providing end-to-end analyses that rigorously explain its advantages in concrete applications. We take a step by studying the effectiveness of a simplified variant of Muon through two case studies: matrix factorization, and in-context learning of linear transformers. For both problems, we prove that simplified Muon converges linearly with iteration complexities independent of the relevant condition number, provably outperforming gradient descent and Adam. Our analysis reveals that the Muon dynamics decouple into a collection of independent scalar sequences in the spectral domain, each exhibiting similar convergence behavior. Our theory formalizes the preconditioning effect induced by spectral orthogonalization, offering insight into Muon's effectiveness in these matrix optimization problems and potentially beyond. --- ### Convolutional Surrogate for 3D Discrete Fracture-Matrix Tensor Upscaling Authors: Martin Špetlík, Jan Březina Published: 2026-01-19 Categories: cs.LG, math.NA arXiv: https://arxiv.org/abs/2604.02335v1 Modeling groundwater flow in three-dimensional fractured crystalline media requires accounting for strong spatial heterogeneity induced by fractures. Fine-scale discrete fracture-matrix (DFM) simulations can capture this complexity but are computationally expensive, especially when repeated evaluations are needed. To address this, we aim to employ a multilevel Monte Carlo (MLMC) framework in which numerical homogenization is used to upscale sub-resolution fracture effects when transitioning between accuracy levels. To reduce the cost of conventional 3D numerical homogenization, we develop a surrogate model that predicts the equivalent hydraulic conductivity tensor Keq from a voxelized 3D domain representing tensor-valued random fields of matrix and fracture conductivities. Fracture size, orientation, and aperture are sampled from distributions informed by natural observations. The surrogate architecture combines a 3D convolutional neural network with feed-forward layers, enabling it to capture both local spatial features and global interactions. Three surrogates are trained on data generated by DFM simulations, each corresponding to a different fracture-to-matrix conductivity contrast. Performance is evaluated across a wide range of fracture network parameters and matrix-field correlation lengths. The trained models achieve high accuracy, with normalized root-mean-square errors below 0.22 across most test cases. Practical applicability is demonstrated by comparing numerically homogenized conductivities with surrogate predictions in two macro-scale problems: computing equivalent conductivity tensors and predicting outflow from a constrained 3D domain. In both cases, surrogate-based upscaling preserves accuracy while substantially reducing computational cost, achieving speedups exceeding 100x when inference is performed on a GPU. --- ### ReCQR: Incorporating conversational query rewriting to improve Multimodal Image Retrieval Authors: Yuan Hu, ZhiYu Cao, PeiFeng Li, QiaoMing Zhu Published: 2026-01-19 Categories: cs.IR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2603.26669v1 With the rise of multimodal learning, image retrieval plays a crucial role in connecting visual information with natural language queries. Existing image retrievers struggle with processing long texts and handling unclear user expressions. To address these issues, we introduce the conversational query rewriting (CQR) task into the image retrieval domain and construct a dedicated multi-turn dialogue query rewriting dataset. Built on full dialogue histories, CQR rewrites users' final queries into concise, semantically complete ones that are better suited for retrieval. Specifically, We first leverage Large Language Models (LLMs) to generate rewritten candidates at scale and employ an LLM-as-Judge mechanism combined with manual review to curate approximately 7,000 high-quality multimodal dialogues, forming the ReCQR dataset. Then We benchmark several SOTA multimodal models on the ReCQR dataset to assess their performance on image retrieval. Experimental results demonstrate that CQR not only significantly enhances the accuracy of traditional image retrieval models, but also provides new directions and insights for modeling user queries in multimodal systems. --- ### Barriers to Discrete Reasoning with Transformers: A Survey Across Depth, Exactness, and Bandwidth Authors: Michelle Yuan, Weiyi Sun, Amir H. Rezaeian, Jyotika Singh, Sandip Ghoshal, Yao-Ting Wang, Miguel Ballesteros, Yassine Benajiba Published: 2026-01-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.11175v1 Transformers have become the foundational architecture for a broad spectrum of sequence modeling applications, underpinning state-of-the-art systems in natural language processing, vision, and beyond. However, their theoretical limitations in discrete reasoning tasks, such as arithmetic, logical inference, and algorithmic composition, remain a critical open problem. In this survey, we synthesize recent studies from three theoretical perspectives: circuit complexity, approximation theory, and communication complexity, to clarify the structural and computational barriers that transformers face when performing symbolic computations. By connecting these established theoretical frameworks, we provide an accessible and unified account of why current transformer architectures struggle to implement exact discrete algorithms, even as they excel at pattern matching and interpolation. We review key definitions, seminal results, and illustrative examples, highlighting challenges such as depth constraints, difficulty approximating discontinuities, and bottlenecks in inter-token communication. Finally, we discuss implications for model design and suggest promising directions for overcoming these foundational limitations. --- ### The Script Tax: Measuring Tokenization-Driven Efficiency and Latency Disparities in Multilingual Language Models Authors: Aradhya Dixit, Shreem Dixit Published: 2026-01-19 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.11174v1 Pretrained multilingual language models are often assumed to be script-agnostic, yet their tokenizers can impose systematic costs on certain writing systems. We quantify this script tax by comparing two orthographic variants with identical linguistic content. Across mBERT and XLM-R, the higher-fragmentation orthography shows a ~3.4x increase in fertility (6.73-6.85 vs. 2.10-2.35 tokens/word), leading to a 16.5x inference slowdown (0.23 vs. 3.8 sentences/second) on identical hardware. Using bits per character (BPC) to avoid the "NLL paradox" from subword fragmentation, we find a substantial increase in information cost: +19.7% for mBERT (8.06->9.65) and +47.1% for XLM-R (12.19->17.94). A round-trip conversion check (CER_rt=0.31) suggests these gaps reflect orthography-conditioned processing rather than mapping noise. Our results highlight tokenization as a key source of inequity in multilingual NLP and motivate script-aware tokenization and pretraining. --- ### Author-in-the-Loop Response Generation and Evaluation: Integrating Author Expertise and Intent in Responses to Peer Review Authors: Qian Ruan, Iryna Gurevych Published: 2026-01-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.11173v3 Author response (rebuttal) writing is a critical stage of scientific peer review that demands substantial author effort. In practice, authors possess domain expertise, author-only information, and response strategies - concrete forms of author expertise and intent - and seek NLP assistance that integrates these signals into author response generation (ARG). Yet this author-in-the-loop paradigm lacks formal NLP formulation and systematic study: no dataset provides fine-grained author signals, existing ARG work lacks author inputs and controls, and no evaluation measures response reflection of author signals and effectiveness in addressing reviewer concerns. To fill these gaps, we introduce (i) Re3Align, the first large-scale dataset of aligned review-response-revision triplets, where revisions proxy author signals; (ii) REspGen, an author-in-the-loop ARG framework supporting flexible author input, multi-attribute control, and evaluation-guided refinement; and (iii) REspEval, a comprehensive evaluation suite with 20+ metrics spanning input utilization, controllability, response quality, and discourse. Experiments with SOTA LLMs demonstrate the benefits of author input and evaluation-guided refinement, the impact of input specificity on response quality, and controllability-quality trade-offs. We release our dataset, generation and evaluation tools. --- ### Synthesizing the Virtual Advocate: A Multi-Persona Speech Generation Framework for Diverse Linguistic Jurisdictions in Indic Languages Authors: Aniket Deroy Published: 2026-01-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.11172v1 Legal advocacy requires a unique combination of authoritative tone, rhythmic pausing for emphasis, and emotional intelligence. This study investigates the performance of the Gemini 2.5 Flash TTS and Gemini 2.5 Pro TTS models in generating synthetic courtroom speeches across five Indic languages: Tamil, Telugu, Bengali, Hindi, and Gujarati. We propose a prompting framework that utilizes Gemini 2.5s native support for 5 languages and its context-aware pacing to produce distinct advocate personas. The evolution of Large Language Models (LLMs) has shifted the focus of TexttoSpeech (TTS) technology from basic intelligibility to context-aware, expressive synthesis. In the legal domain, synthetic speech must convey authority and a specific professional persona a task that becomes significantly more complex in the linguistically diverse landscape of India. The models exhibit a "monotone authority," excelling at procedural information delivery but struggling with the dynamic vocal modulation and emotive gravitas required for persuasive advocacy. Performance dips in Bengali and Gujarati further highlight phonological frontiers for future refinement. This research underscores the readiness of multilingual TTS for procedural legal tasks while identifying the remaining challenges in replicating the persuasive artistry of human legal discourse. The code is available at-https://github.com/naturenurtureelite/Synthesizing-the-Virtual-Advocate/tree/main --- ### A Language-Guided Bayesian Optimization for Efficient LoRA Hyperparameter Search Authors: Baek Seong-Eun, Lee Jung-Mok, Kim Sung-Bin, Tae-Hyun Oh Published: 2026-01-19 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.11171v2 Fine-tuning Large Language Models (LLMs) with Low-Rank Adaptation (LoRA) offers a resource-efficient way to personalize or specialize. However, LoRA is highly sensitive to hyperparameter choices, and exhaustive hyperparameter search is computationally expensive. To address this, we propose a Bayesian Optimization (BO) framework that leverages the domain knowledge of pre-trained LLMs to efficiently search for LoRA hyperparameters. Our approach repurposes a pre-trained LLM as a discrete-to-continuous mapping module to link hyperparameters and their domain knowledge to a continuous vector space, where BO is conducted. We design and control the mapping via language prompting, providing a domain-aware textual prompt that describes the relationships among hyperparameters and their respective roles. This allows us to explicitly inject domain knowledge about LoRA into the LLM in natural language. We also introduce an additional learnable token to capture residual information that is difficult to describe linguistically in the prompt. This aids BO to sample more high-performing hyperparameters. In addition, by leveraging the strong correlation observed between the performance obtained from full and subset training datasets in LoRA training regimes, we introduce proxy training and evaluation using a data subset. This significantly improves the efficiency of our method. We demonstrate that our hyperparameter, discovered with only about 30 iterations, achieves more than 20% performance improvement over standard hyperparameters found from about 45,000 combinations. Project page: https://baekseongeun.github.io/lora-bo/ --- ### PRIME: Policy-Reinforced Iterative Multi-agent Execution for Algorithmic Reasoning in Large Language Models Authors: Jiawei Xu, Zhenyu Yu, Ziqian Bi, Minh Duc Pham, Xiaoyi Qu, Danyang Zhang Published: 2026-01-19 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.11170v1 Large language models have demonstrated remarkable capabilities across diverse reasoning tasks, yet their performance on algorithmic reasoning remains limited. To handle this limitation, we propose PRIME (Policy-Reinforced Iterative Multi-agent Execution), a framework comprising three specialized agents, an executor for step-by-step reasoning, a verifier for constraint checking, and a coordinator for backtracking control, optimized through group relative policy optimization. For comprehensive evaluation, we introduce PRIME-Bench, the largest algorithmic reasoning benchmark to date, comprising 86 tasks across 12 categories with 51,600 instances. Tasks span sorting algorithms, graph and tree structures, automata and state machines, symbolic reasoning, and constraint-based puzzles, with execution traces reaching over one million steps. Compared to baseline approach, PRIME improves average accuracy from 26.8% to 93.8%, a 250% relative gain. The largest improvements occur on tasks requiring sustained state tracking, with Turing machine simulation improving from 9% to 92% and long division from 16% to 94%. Ablation studies identify iterative verification as the primary contributor, preventing the error propagation that causes baseline approaches to fail catastrophically. Analysis across model scales (8B-120B parameters) reveals that smaller models benefit disproportionately, achieving accuracy comparable to models 8x larger. --- ### Disentangling Direction and Magnitude in Transformer Representations: A Double Dissociation Through L2-Matched Perturbation Analysis Authors: Mangadoddi Srikar Vardhan, Lekkala Sai Teja Published: 2026-01-19 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.11169v1 Transformer hidden states encode information as high-dimensional vectors, yet whether direction (orientation in representational space) and magnitude (vector norm) serve distinct functional roles remains unclear. Studying Pythia-family models, we discover a striking cross-over dissociation: angular perturbations cause up to 42.9 more damage to language modeling loss, while magnitude perturbations cause disproportionately more damage to syntactic processing (20.4% vs.1.6% accuracy drop on subject-verb agreement).This finding is enabled by L2-matched perturbation analysis, a methodology ensuring that an gular and magnitude perturbations achieve identical Euclidean displacements. Causal intervention reveals that angular damage flows substantially through the attention pathways (28.4% loss recovery via attention repair), while magnitude damage flows partly through the LayerNorm pathways(29.9% recovery via LayerNorm repair). These patterns replicate across scales within the Pythia architecture family. These findings provide evidence that direction and magnitude support partially distinct computational roles in LayerNorm based architectures. The direction preferentially affects attentional routing, while magnitude modulates processing intensity for fine-grained syntactic judgments. We find different patterns in RMSNorm-based architectures, suggesting that the dissociation depends on architectural choices. Our results refine the linear representation hypothesis and have implications for model editing and interpretability research --- ### Enhancing SDG-Text Classification with Combinatorial Fusion Analysis and Generative AI Authors: Jingyan Xu, Marcelo L. LaFleur, Christina Schweikert, D. Frank Hsu Published: 2026-01-19 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.11168v1 (Natural Language Processing) NLP techniques such as text classification and topic discovery are very useful in many application areas including information retrieval, knowledge discovery, policy formulation, and decision-making. However, it remains a challenging problem in cases where the categories are unavailable, difficult to differentiate, or are interrelated. Social analysis with human context is an area that can benefit from text classification, as it relies substantially on text data. The focus of this paper is to enhance the classification of text according to the UN's Sustainable Development Goals (SDGs) by collecting and combining intelligence from multiple models. Combinatorial Fusion Analysis (CFA), a system fusion paradigm using a rank-score characteristic (RSC) function and cognitive diversity (CD), has been used to enhance classifier methods by combining a set of relatively good and mutually diverse classification models. We use a generative AI model to generate synthetic data for model training and then apply CFA to this classification task. The CFA technique achieves 96.73% performance, outperforming the best individual model. We compare the outcomes with those obtained from human domain experts. It is demonstrated that combining intelligence from multiple ML/AI models using CFA and getting input from human experts can, not only complement, but also enhance each other. --- ### Analyzing Diffusion and Autoregressive Vision Language Models in Multimodal Embedding Space Authors: Zihang Wang, Siyue Zhang, Yilun Zhao, Jingyi Yang, Tingyu Song, Anh Tuan Luu, Chen Zhao Published: 2026-01-19 Categories: cs.MM, cs.AI, cs.CL, cs.CV arXiv: https://arxiv.org/abs/2602.06056v1 Embedding models are a fundamental component of modern AI systems such as semantic search and retrieval-augmented generation. Recent advances in large foundation models have substantially accelerated the development of embedding models, including those based on Large Language Models (LLMs), Vision Language Models (VLMs), and Multimodal LLMs. More recently, Large Diffusion Language Models (dLLMs) and Multimodal dLLMs have emerged as competitive alternatives to autoregressive models, offering advantages such as bidirectional attention and parallel generation. This progress naturally raises a critical yet unexplored question: can Multimodal dLLMs serve as effective multimodal embedding models? To answer this, we present the first systematic study of converting Multimodal dLLMs into embedding models. We evaluate state-of-the-art Multimodal dLLMs and Autoregressive VLMs across three categories of embedding tasks: classification, visual question answering, and information retrieval. Our results show that Multimodal dLLM embeddings generally underperform their autoregressive VLM counterparts. The stronger diffusion-based model, LaViDa, lags by only 3.5 points on classification, 2.5 points on VQA, and 4.4 points on retrieval tasks, whereas the other diffusion-based model, MMaDA, exhibits substantially larger performance gaps, exceeding 20 points across all tasks. Further analysis reveals insufficient image-text alignment in diffusion-based models, accounting for the observed limitations in their embedding performance. --- ### Augmenting Parameter-Efficient Pre-trained Language Models with Large Language Models Authors: Saurabh Anand, Shubham Malaviya, Manish Shukla, Sachin Lodha Published: 2026-01-19 Categories: cs.LG, cs.CR arXiv: https://arxiv.org/abs/2602.02501v1 Training AI models in cybersecurity with help of vast datasets offers significant opportunities to mimic real-world behaviors effectively. However, challenges like data drift and scarcity of labelled data lead to frequent updates of models and the risk of overfitting. To address these challenges, we used parameter-efficient fine-tuning techniques for pre-trained language models wherein we combine compacters with various layer freezing strategies. To enhance the capabilities of these pre-trained language models, in this work we introduce two strategies that use large language models. In the first strategy, we utilize large language models as data-labelling tools wherein they generate labels for unlabeled data. In the second strategy, large language modes are utilized as fallback mechanisms for predictions having low confidence scores. We perform comprehensive experimental analysis on the proposed strategies on different downstream tasks specific to cybersecurity domain. We empirically demonstrate that by combining parameter-efficient pre-trained models with large language models, we can improve the reliability and robustness of models, making them more suitable for real-world cybersecurity applications. --- ### Scalable and Secure AI Inference in Healthcare: A Comparative Benchmarking of FastAPI and Triton Inference Server on Kubernetes Authors: Ratul Ali Published: 2026-01-19 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.00053v1 Efficient and scalable deployment of machine learning (ML) models is a prerequisite for modern production environments, particularly within regulated domains such as healthcare and pharmaceuticals. In these settings, systems must balance competing requirements, including minimizing inference latency for real-time clinical decision support, maximizing throughput for batch processing of medical records, and ensuring strict adherence to data privacy standards such as HIPAA. This paper presents a rigorous benchmarking analysis comparing two prominent deployment paradigms: a lightweight, Python-based REST service using FastAPI, and a specialized, high-performance serving engine, NVIDIA Triton Inference Server. Leveraging a reference architecture for healthcare AI, we deployed a DistilBERT sentiment analysis model on Kubernetes to measure median (p50) and tail (p95) latency, as well as throughput, under controlled experimental conditions. Our results indicate a distinct trade-off. While FastAPI provides lower overhead for single-request workloads with a p50 latency of 22 ms, Triton achieves superior scalability through dynamic batching, delivering a throughput of 780 requests per second on a single NVIDIA T4 GPU, nearly double that of the baseline. Furthermore, we evaluate a hybrid architectural approach that utilizes FastAPI as a secure gateway for protected health information de-identification and Triton for backend inference. This study validates the hybrid model as a best practice for enterprise clinical AI and offers a blueprint for secure, high-availability deployments. --- ### AI-assisted Protocol Information Extraction For Improved Accuracy and Efficiency in Clinical Trial Workflows Authors: Ramtin Babaeipour, François Charest, Madison Wright Published: 2026-01-19 Categories: cs.IR, cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2602.00052v2 Increasing clinical trial protocol complexity, amendments, and challenges around knowledge management create significant burden for trial teams. Structuring protocol content into standard formats has the potential to improve efficiency, support documentation quality, and strengthen compliance. We evaluate an Artificial Intelligence (AI) system using generative LLMs with Retrieval-Augmented Generation (RAG) for automated clinical trial protocol information extraction. We compare the extraction accuracy of our clinical-trial-specific RAG process against that of publicly available (standalone) LLMs. We also assess the operational impact of AI-assistance on simulated extraction Clinical Research Coordinator (CRC) workflows. Our RAG process shows higher extraction accuracy (89.0%) than standalone LLMs with fine-tuned prompts (62.6%) against expert-supported reference annotations. In simulated extraction workflows, AI-assisted tasks are completed 40% faster, are rated as less cognitively demanding and are strongly preferred by users. While expert oversight remains essential, this suggests that AI-assisted extraction can enable protocol intelligence at scale, motivating the integration of similar methodologies into real-world clinical workflows to further validate its impact on feasibility, study start-up, and post-activation monitoring. --- ### Distributional Reinforcement Learning for Condition-Based Maintenance of Multi-Pump Equipment Authors: Takato Yasuno Published: 2026-01-19 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.00051v1 Condition-Based Maintenance (CBM) signifies a paradigm shift from reactive to proactive equipment management strategies in modern industrial systems. Conventional time-based maintenance schedules frequently engender superfluous expenditures and unanticipated equipment failures. In contrast, CBM utilizes real-time equipment condition data to enhance maintenance timing and optimize resource allocation. The present paper proposes a novel distributional reinforcement learning approach for multi-equipment CBM using Quantile Regression Deep Q-Networks (QR-DQN) with aging factor integration. The methodology employed in this study encompasses the concurrent administration of multiple pump units through three strategic scenarios. The implementation of safety-first, balanced, and cost-efficient approaches is imperative. Comprehensive experimental validation over 3,000 training episodes demonstrates significant performance improvements across all strategies. The Safety-First strategy demonstrates superior cost efficiency, with a return on investment (ROI) of 3.91, yielding 152\% better performance than alternatives while requiring only 31\% higher investment. The system exhibits 95.66\% operational stability and immediate applicability to industrial environments. --- ### Exploring the Interpretability of Forecasting Models for Energy Balancing Market Authors: Oskar Våle, Shiliang Zhang, Sabita Maharjan, Gro Klæboe Published: 2026-01-19 Categories: q-fin.ST, cs.LG, q-fin.GN arXiv: https://arxiv.org/abs/2602.00049v1 The balancing market in the energy sector plays a critical role in physically and financially balancing the supply and demand. Modeling dynamics in the balancing market can provide valuable insights and prognosis for power grid stability and secure energy supply. While complex machine learning models can achieve high accuracy, their black-box nature severely limits the model interpretability. In this paper, we explore the trade-off between model accuracy and interpretability for the energy balancing market. Particularly, we take the example of forecasting manual frequency restoration reserve (mFRR) activation price in the balancing market using real market data from different energy price zones. We explore the interpretability of mFRR forecasting using two models: extreme gradient boosting (XGBoost) machine and explainable boosting machine (EBM). We also integrate the two models, and we benchmark all the models against a baseline naive model. Our results show that EBM provides forecasting accuracy comparable to XGBoost while yielding a considerable level of interpretability. Our analysis also underscores the challenge of accurately predicting the mFRR price for the instances when the activation price deviates significantly from the spot price. Importantly, EBM's interpretability features reveal insights into non-linear mFRR price drivers and regional market dynamics. Our study demonstrates that EBM is a viable and valuable interpretable alternative to complex black-box AI models in the forecast for the balancing market. --- ### Quantum Circuit-Based Learning Models: Bridging Quantum Computing and Machine Learning Authors: Fan Fan, Yilei Shi, Mihai Datcu, Bertrand Le Saux, Luigi Iapichino, Francesca Bovolo, Silvia Liberata Ullo, Xiao Xiang Zhu Published: 2026-01-19 Categories: quant-ph, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.00048v1 Machine Learning (ML) has been widely applied across numerous domains due to its ability to automatically identify informative patterns from data for various tasks. The availability of large-scale data and advanced computational power enables the development of sophisticated models and training strategies, leading to state-of-the-art performance, but it also introduces substantial challenges. Quantum Computing (QC), which exploits quantum mechanisms for computation, has attracted growing attention and significant global investment as it may address these challenges. Consequently, Quantum Machine Learning (QML), the integration of these two fields, has received increasing interest, with a notable rise in related studies in recent years. We are motivated to review these existing contributions regarding quantum circuit-based learning models for classical data analysis and highlight the identified potentials and challenges of this technique. Specifically, we focus not only on QML models, both kernel-based and neural network-based, but also on recent explorations of their integration with classical machine learning layers within hybrid frameworks. Moreover, we examine both theoretical analysis and empirical findings to better understand their capabilities, and we also discuss the efforts on noise-resilient and hardware-efficient QML that could enhance its practicality under current hardware limitations. In addition, we cover several emerging paradigms for advanced quantum circuit design and highlight the adaptability of QML across representative application domains. This study aims to provide an overview of the contributions made to bridge quantum computing and machine learning, offering insights and guidance to support its future development and pave the way for broader adoption in the coming years. --- ### Lightweight Edge Learning via Dataset Pruning Authors: Laha Ale, Hu Luo, Mingsheng Cao, Shichao Li, Huanlai Xing, Haifeng Sun Published: 2026-01-19 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2602.00047v1 Edge learning facilitates ubiquitous intelligence by enabling model training and adaptation directly on data-generating devices, thereby mitigating privacy risks and communication latency. However, the high computational and energy overhead of on-device training hinders its deployment on battery-powered mobile systems with strict thermal and memory budgets. While prior research has extensively optimized model architectures for efficient inference, the training phase remains bottlenecked by the processing of massive, often redundant, local datasets. In this work, we propose a data-centric optimization framework that leverages dataset pruning to achieve resource-efficient edge learning. Unlike standard methods that process all available data, our approach constructs compact, highly informative training subsets via a lightweight, on-device importance evaluation. Specifically, we utilize average loss statistics derived from a truncated warm-up phase to rank sample importance, deterministically retaining only the most critical data points under a dynamic pruning ratio. This mechanism is model-agnostic and operates locally without inter-device communication. Extensive experiments on standard image classification benchmarks demonstrate that our framework achieves a near-linear reduction in training latency and energy consumption proportional to the pruning ratio, with negligible degradation in model accuracy. These results validate dataset pruning as a vital, complementary paradigm for enhancing the sustainability and scalability of learning on resource-constrained mobile edge devices. --- ### Extending Beacon to Hindi: Cultural Adaptation Drives Cross-Lingual Sycophancy Authors: Sarthak Sattigeri Published: 2026-01-19 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2602.00046v1 Sycophancy, the tendency of language models to prioritize agreement with user preferences over principled reasoning, has been identified as a persistent alignment failure in English-language evaluations. However, it remains unclear whether such diagnostics generalize across languages and cultural contexts. We extend the Beacon single-turn forced-choice sycophancy diagnostic to Hindi through a controlled three-condition design: English original, Hindi literal translation, and Hindi culturally adapted prompts. We evaluate four open-weight instruction-tuned models on 50 prompts per condition, enabling separation of language encoding effects from cultural adaptation effects. Across all models, sycophancy rates are consistently higher for culturally adapted Hindi prompts than for English, with absolute differences ranging from 12.0 to 16.0 percentage points. A decomposition on Qwen 2.5-Coder-7B shows that cultural adaptation (delta = 14.0%, 95% CI: (4.0%, 26.0%)) accounts for the majority of this gap, while language encoding contributes minimally (delta = 2.0%, 95% CI: (0.0%, 6.0%)). Category-level analysis reveals that advice prompts exhibit the largest cross-lingual differences (20-25 percentage points), achieving statistical significance in two of four models. These findings indicate that alignment behaviors measured in English may not transfer uniformly across languages and that culturally grounded prompt framing plays a substantial role. We release all datasets and evaluation code to support replication and extension. --- ### When LLMs Imagine People: A Human-Centered Persona Brainstorm Audit for Bias and Fairness in Creative Applications Authors: Hongliu Cao, Eoin Thomas, Rodrigo Acuna Agost Published: 2026-01-19 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2602.00044v2 Large Language Models (LLMs) used in creative workflows can reinforce stereotypes and perpetuate inequities, making fairness auditing essential. Existing methods rely on constrained tasks and fixed benchmarks, leaving open-ended creative outputs unexamined. We introduce the Persona Brainstorm Audit (PBA), a scalable and easy to extend auditing method for bias detection across multiple intersecting identity and social roles in open-ended persona generation. PBA quantifies bias using degree-of-freedom-aware normalized Cramér's V, producing interpretable severity labels that enable fair comparison across models and dimensions. Applying PBA to 12 LLMs (120,000 personas, 16 bias dimensions), we find that bias evolves nonlinearly across model generations: larger and newer models are not consistently fairer, and biases that initially decrease can resurface in later releases. Intersectional analysis reveals disparities hidden by single-axis metrics, where dimensions appearing fair individually can exhibit high bias in combination. Robustness analyses show PBA remains stable under varying sample sizes, role-playing prompts, and debiasing prompts, establishing its reliability for fairness auditing in LLMs. --- ### JSR-GFNet: Jamming-to-Signal Ratio-Aware Dynamic Gating for Interference Classification in future Cognitive Global Navigation Satellite Systems Authors: Zhihan Zeng, Hongyuan Shu, Kaihe Wang, Lu Chen, Amir Hussian, Yanjun Huang, Junchu Zhao, Yue Xiu, Zhongpei Zhang Published: 2026-01-19 Categories: eess.SP, cs.AI arXiv: https://arxiv.org/abs/2602.00042v1 The transition toward cognitive global navigation satellite system (GNSS) receivers requires accurate interference classification to trigger adaptive mitigation strategies. However, conventional methods relying on Time-Frequency Analysis (TFA) and Convolutional Neural Networks (CNNs) face two fundamental limitations: severe performance degradation in low Jamming-to-Signal Ratio (JSR) regimes due to noise obscuration, and ``feature degeneracy'' caused by the loss of phase information in magnitude-only spectrograms. Consequently, spectrally similar signals -- such as high-order Quadrature Amplitude Modulation versus Band-Limited Gaussian Noise -- become indistinguishable. To overcome these challenges, this paper proposes the \textbf{JSR-Guided Fusion Network (JSR-GFNet)}. This multi-modal architecture combines phase-sensitive complex In-Phase/Quadrature (IQ) samples with Short-Time Fourier Transform (STFT) spectrograms. Central to this framework is a physics-inspired dynamic gating mechanism driven by statistical signal descriptors. Acting as a conditional controller, it autonomously estimates signal reliability to dynamically reweight the contributions of a Complex-Valued ResNet (IQ stream) and an EfficientNet backbone (STFT stream). To validate the model, we introduce the Comprehensive GNSS Interference (CGI-21) dataset, simulating 21 jamming categories including software-defined waveforms from aerial platforms. Extensive experiments demonstrate that JSR-GFNet achieves higher accuracy across the full 10--50 dB JSR spectrum. Notably, interpretability analysis confirms that the model learns a physically intuitive strategy: prioritizing spectral energy integration in noise-limited regimes while shifting focus to phase precision in high-SNR scenarios to resolve modulation ambiguities. This framework provides a robust solution for next-generation aerospace navigation security. --- ### Student Perceptions of Large Language Models Use in Self-Reflection and Design Critique in Architecture Studio Authors: Juan David Salazar Rodriguez, Sam Conrad Joyce, Nachamma Sockalingam, Khoo Eng Tat, Julfendi Published: 2026-01-19 Categories: cs.CY, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2602.00041v2 This study investigates the integration of Large Language Models (LLMs) into the feedback mechanisms of the architectural design studio, shifting the focus from generative production to reflective pedagogy. Employing a mixed-methods approach with surveys and semi structured interviews with 22 architecture students at the Singapore University of Technology and De-sign, the research analyzes student perceptions across three distinct feed-back domains: self-reflection, peer critique, and professor-led reviews. The findings reveal that students engage with LLMs not as authoritative in-structors, but as collaborative "cognitive mirrors" that scaffold critical thinking. In self-directed learning, LLMs help structure thoughts and over-come the "blank page" problem, though they are limited by a lack of contex-tual nuance. In peer critiques, the technology serves as a neutral mediator, mitigating social anxiety and the "fear of offending". Furthermore, in high-stakes professor-led juries, students utilize LLMs primarily as post-critique synthesis engines to manage cognitive overload and translate ab-stract academic discourse into actionable design iterations. --- ### Enhancing few-shot time series forecasting with LLM-guided diffusion Authors: Haonan Shi, Dehua Shuai, Liming Wang, Xiyang Liu, Long Tian Published: 2026-01-19 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2602.00040v1 Time series forecasting in specialized domains is often constrained by limited data availability, where conventional models typically require large-scale datasets to effectively capture underlying temporal dynamics. To tackle this few-shot challenge, we propose LTSM-DIFF (Large-scale Temporal Sequential Memory with Diffusion), a novel learning framework that integrates the expressive power of large language models with the generative capability of diffusion models. Specifically, the LTSM module is fine-tuned and employed as a temporal memory mechanism, extracting rich sequential representations even under data-scarce conditions. These representations are then utilized as conditional guidance for a joint probability diffusion process, enabling refined modeling of complex temporal patterns. This design allows knowledge transfer from the language domain to time series tasks, substantially enhancing both generalization and robustness. Extensive experiments across diverse benchmarks demonstrate that LTSM-DIFF consistently achieves state-of-the-art performance in data-rich scenarios, while also delivering significant improvements in few-shot forecasting. Our work establishes a new paradigm for time series analysis under data scarcity. --- ### LSSF: Safety Alignment for Large Language Models through Low-Rank Safety Subspace Fusion Authors: Guanghao Zhou, Panjia Qiu, Cen Chen, Hongyu Li, Mingyuan Chu, Xin Zhang, Jun Zhou Published: 2026-01-19 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2602.00038v1 The safety mechanisms of large language models (LLMs) exhibit notable fragility, as even fine-tuning on datasets without harmful content may still undermine their safety capabilities. Meanwhile, existing safety alignment methods predominantly rely on the fine-tuning process, which inadvertently leads to the increased complexity and computational resources required. To address these issues, we introduce LSSF, a novel safety re-alignment framework with \underline{L}ow-Rank \underline{S}afety \underline{S}ubspace \underline{F}usion. Our proposed method exploits the low-rank characteristics of safety information in LLMs by constructing a low-rank projection matrix to extract the principal components of safety vectors. Notably, this projection matrix represents the low-rank safety subspace of the LLMs, which we have observed to remain stable during fine-tuning process and is isolated from the model's general capabilities. These principal components are used to effectively restore safety alignment when combined with fine-tuned LLMs through linear arithmetic. Additionally, to account for the varying encoding densities of safety information across different layers of LLMs, we propose a novel metric called safety singular value entropy. This metric quantifies the encoding density and allows for the dynamic computation of the safety-critical rank for each safety vector. Extensive experiments demonstrate that our proposed post-hoc alignment method can effectively restore the safety alignment of fine-tuned models with minimal impact on their performance in downstream tasks. --- ### Bitcoin Price Prediction using Machine Learning and Combinatorial Fusion Analysis Authors: Yuanhong Wu, Wei Ye, Jingyan Xu, D. Frank Hsu Published: 2026-01-19 Categories: q-fin.ST, cs.AI, cs.CE, cs.LG arXiv: https://arxiv.org/abs/2602.00037v2 In this work, we propose to apply a new model fusion and learning paradigm, known as Combinatorial Fusion Analysis (CFA), to the field of Bitcoin price prediction. Price prediction of financial product has always been a big topic in finance, as the successful prediction of the price can yield significant profit. Every machine learning model has its own strength and weakness, which hinders progress toward robustness. CFA has been used to enhance models by leveraging rank-score characteristic (RSC) function and cognitive diversity in the combination of a moderate set of diverse and relatively well-performed models. Our method utilizes both score and rank combinations as well as other weighted combination techniques. Key metrics such as RMSE and MAPE are used to evaluate our methodology performance. Our proposal presents a notable MAPE performance of 0.19\%. The proposed method greatly improves upon individual model performance, as well as outperforms other Bitcoin price prediction models. --- ### In Vino Veritas and Vulnerabilities: Examining LLM Safety via Drunk Language Inducement Authors: Anudeex Shetty, Aditya Joshi, Salil S. Kanhere Published: 2026-01-19 Categories: cs.CL, cs.AI, cs.CR, cs.LG arXiv: https://arxiv.org/abs/2601.22169v1 Humans are susceptible to undesirable behaviours and privacy leaks under the influence of alcohol. This paper investigates drunk language, i.e., text written under the influence of alcohol, as a driver for safety failures in large language models (LLMs). We investigate three mechanisms for inducing drunk language in LLMs: persona-based prompting, causal fine-tuning, and reinforcement-based post-training. When evaluated on 5 LLMs, we observe a higher susceptibility to jailbreaking on JailbreakBench (even in the presence of defences) and privacy leaks on ConfAIde, where both benchmarks are in English, as compared to the base LLMs as well as previously reported approaches. Via a robust combination of manual evaluation and LLM-based evaluators and analysis of error categories, our findings highlight a correspondence between human-intoxicated behaviour, and anthropomorphism in LLMs induced with drunk language. The simplicity and efficiency of our drunk language inducement approaches position them as potential counters for LLM safety tuning, highlighting significant risks to LLM safety. --- ### Drivers, Receivers, and Dynamic Linkages: The Directed Structure of SDG Interdependence, 2000--2024 Authors: Md Muhtasim Munif Fahim, Md Jahid Hasan Imran, Md. Naim Molla, Luknath Debnath, Tonmoy Shil, Ehsanul Bashar Pranto, Md Mostafizur Rahman Likhon, Md Shafin Sanyan Saad, Md. Rezaul Karim Published: 2026-01-19 Categories: stat.AP, cs.LG, econ.EM, stat.ME, stat.ML arXiv: https://arxiv.org/abs/2601.20875v2 Governments with limited fiscal and administrative capacity need to know which Sustainable Development Goals (SDGs) propagate progress through the goal system and how quickly. We map the directed interdependence structure of all seventeen goals using a balanced panel of 114 countries observed annually from 2000 to 2024. The goal series are persistent, trending, and cross-sectionally dependent, so we apply two estimators matched to this regime: a Dumitrescu-Hurlin panel Granger non-causality test, run on first-differenced series, to recover the directed interaction network, and panel local projections with Driscoll-Kraay standard errors to measure the dynamic magnitude of 31 theory-derived indicator linkages. Of 272 directed goal pairs, 84 linkages survive false-discovery control (40 synergies, 44 trade-offs; network density 0.31). Synergies and trade-offs occur at comparable strength, so no single goal behaves as a universal accelerator, and the goal-level hierarchy itself is fragile. Driver-receiver rankings correlate weakly across lag orders and centrality metrics, and under a country bootstrap only two roles are distinguishable from zero: peace and strong institutions as the clearest net receiver, and poverty reduction as the most probable effect-size-weighted driver. The supported linkages are dynamic, accruing over four to five years: sanitation and poverty improvements are the strongest predictors of lower child mortality, and the education-child-health association is corroborated in independent World Development Indicators data across 183 countries. These results caution against rankings-based accelerator policy and support adaptive portfolios built on supported, time-lagged linkages monitored through constituent indicators. --- ### Classifier Calibration at Scale: An Empirical Study of Model-Agnostic Post-Hoc Methods Authors: Valery Manokhin, Daniel Grønhaug Published: 2026-01-19 Categories: cs.LG, stat.AP, stat.ML arXiv: https://arxiv.org/abs/2601.19944v1 We study model-agnostic post-hoc calibration methods intended to improve probabilistic predictions in supervised binary classification on real i.i.d. tabular data, with particular emphasis on conformal and Venn-based approaches that provide distribution-free validity guarantees under exchangeability. We benchmark 21 widely used classifiers, including linear models, SVMs, tree ensembles (CatBoost, XGBoost, LightGBM), and modern tabular neural and foundation models, on binary tasks from the TabArena-v0.1 suite using randomized, stratified five-fold cross-validation with a held-out test fold. Five calibrators; Isotonic regression, Platt scaling, Beta calibration, Venn-Abers predictors, and Pearsonify are trained on a separate calibration split and applied to test predictions. Calibration is evaluated using proper scoring rules (log-loss and Brier score) and diagnostic measures (Spiegelhalter's Z, ECE, and ECI), alongside discrimination (AUC-ROC) and standard classification metrics. Across tasks and architectures, Venn-Abers predictors achieve the largest average reductions in log-loss, followed closely by Beta calibration, while Platt scaling exhibits weaker and less consistent effects. Beta calibration improves log-loss most frequently across tasks, whereas Venn-Abers displays fewer instances of extreme degradation and slightly more instances of extreme improvement. Importantly, we find that commonly used calibration procedures, most notably Platt scaling and isotonic regression, can systematically degrade proper scoring performance for strong modern tabular models. Overall classification performance is often preserved, but calibration effects vary substantially across datasets and architectures, and no method dominates uniformly. In expectation, all methods except Pearsonify slightly increase accuracy, but the effect is marginal, with the largest expected gain about 0.008%. --- ### LLM-Generated or Human-Written? Comparing Review and Non-Review Papers on ArXiv Authors: Yanai Elazar, Maria Antoniak Published: 2026-01-19 Categories: cs.DL, cs.CL, cs.CY arXiv: https://arxiv.org/abs/2601.17036v2 ArXiv recently prohibited the upload of unpublished review papers to its servers in the Computer Science domain, citing a high prevalence of LLM-generated content in these categories. However, this decision was not accompanied by quantitative evidence. In this work, we investigate this claim by measuring the proportion of LLM-generated content in review vs. non-review research papers in recent years. Using two high-quality detection methods, we find a substantial increase in LLM-generated content across both review and non-review papers, with a higher prevalence in review papers. However, when considering the number of LLM-generated papers published in each category, the estimates of non-review LLM-generated papers are almost six times higher. Furthermore, we find that this policy will affect papers in certain domains far more than others, with the CS subdiscipline Computers & Society potentially facing cuts of 50%. Our analysis provides an evidence-based framework for evaluating such policy decisions, and we release our code to facilitate future investigations at: https://github.com/yanaiela/llm-review-arxiv. --- ### Diagnosis Support of Sickle Cell Anemia by Classifying Red Blood Cell Shape in Peripheral Blood Images Authors: Wilkie Delgado-Font, Miriela Escobedo-Nicot, Manuel González-Hidalgo, Silena Herold-Garcia, Antoni Jaume-i-Capó, Arnau Mir Published: 2026-01-19 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2601.17032v1 Red blood cell (RBC) deformation is the consequence of several diseases, including sickle cell anemia, which causes recurring episodes of pain and severe pronounced anemia. Monitoring patients with these diseases involves the observation of peripheral blood samples under a microscope, a time-consuming procedure. Moreover, a specialist is required to perform this technique, and owing to the subjective nature of the observation of isolated RBCs, the error rate is high. In this paper, we propose an automated method for differentially enumerating RBCs that uses peripheral blood smear image analysis. In this method, the objects of interest in the image are segmented using a Chan-Vese active contour model. An analysis is then performed to classify the RBCs, also called erythrocytes, as normal or elongated or having other deformations, using the basic shape analysis descriptors: circular shape factor (CSF) and elliptical shape factor (ESF). To analyze cells that become partially occluded in a cluster during sample preparation, an elliptical adjustment is performed to allow the analysis of erythrocytes with discoidal and elongated shapes. The images of patient blood samples used in the study were acquired by a clinical laboratory specialist in the Special Hematology Department of the ``Dr. Juan Bruno Zayas'' General Hospital in Santiago de Cuba. A comparison of the results obtained by the proposed method in our experiments with those obtained by some state-of-the-art methods showed that the proposed method is superior for the diagnosis of sickle cell anemia. This superiority is achieved for evidenced by the obtained F-measure value (0.97 for normal cells and 0.95 for elongated ones) and several overall multiclass performance measures. The results achieved by the proposed method are suitable for the purpose of clinical treatment and diagnostic support of sickle cell anemia. --- ### Data-Efficient Meningioma Segmentation via Implicit Spatiotemporal Mixing and Sim2Real Semantic Injection Authors: Yunhao Xu, Fuquan Zong, Yexuan Xing, Chulong Zhang, Guang Yang, Shilong Yang, Xiaokun Liang, Juan Yu Published: 2026-01-19 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2601.17031v1 The performance of medical image segmentation is increasingly defined by the efficiency of data utilization rather than merely the volume of raw data. Accurate segmentation, particularly for complex pathologies like meningiomas, demands that models fully exploit the latent information within limited high-quality annotations. To maximize the value of existing datasets, we propose a novel dual-augmentation framework that synergistically integrates spatial manifold expansion and semantic object injection. Specifically, we leverage Implicit Neural Representations (INR) to model continuous velocity fields. Unlike previous methods, we perform linear mixing on the integrated deformation fields, enabling the efficient generation of anatomically plausible variations by interpolating within the deformation space. This approach allows for the extensive exploration of structural diversity from a small set of anchors. Furthermore, we introduce a Sim2Real lesion injection module. This module constructs a high-fidelity simulation domain by transplanting lesion textures into healthy anatomical backgrounds, effectively bridging the gap between synthetic augmentation and real-world pathology. Comprehensive experiments on a hybrid dataset demonstrate that our framework significantly enhances the data efficiency and robustness of state-of-the-art models, including nnU-Net and U-Mamba, offering a potent strategy for high-performance medical image analysis with limited annotation budgets. --- ### ECGomics: An Open Platform for AI-ECG Digital Biomarker Discovery Authors: Deyun Zhang, Jun Li, Shijia Geng, Yue Wang, Shijie Chen, Sumei Fan, Qinghao Zha, Shenda Hong Published: 2026-01-19 Categories: q-bio.QM, cs.AI arXiv: https://arxiv.org/abs/2601.15326v1 Background: Conventional electrocardiogram (ECG) analysis faces a persistent dichotomy: expert-driven features ensure interpretability but lack sensitivity to latent patterns, while deep learning offers high accuracy but functions as a black box with high data dependency. We introduce ECGomics, a systematic paradigm and open-source platform for the multidimensional deconstruction of cardiac signals into digital biomarker. Methods: Inspired by the taxonomic rigor of genomics, ECGomics deconstructs cardiac activity across four dimensions: Structural, Intensity, Functional, and Comparative. This taxonomy synergizes expert-defined morphological rules with data-driven latent representations, effectively bridging the gap between handcrafted features and deep learning embeddings. Results: We operationalized this framework into a scalable ecosystem consisting of a web-based research platform and a mobile-integrated solution (https://github.com/PKUDigitalHealth/ECGomics). The web platform facilitates high-throughput analysis via precision parameter configuration, high-fidelity data ingestion, and 12-lead visualization, allowing for the systematic extraction of biomarkers across the four ECGomics dimensions. Complementarily, the mobile interface, integrated with portable sensors and a cloud-based engine, enables real-time signal acquisition and near-instantaneous delivery of structured diagnostic reports. This dual-interface architecture successfully transitions ECGomics from theoretical discovery to decentralized, real-world health management, ensuring professional-grade monitoring in diverse clinical and home-based settings. Conclusion: ECGomics harmonizes diagnostic precision, interpretability, and data efficiency. By providing a deployable software ecosystem, this paradigm establishes a robust foundation for digital biomarker discovery and personalized cardiovascular medicine. --- ### Tracing the Data Trail: A Survey of Data Provenance, Transparency and Traceability in LLMs Authors: Richard Hohensinner, Belgin Mutlu, Inti Gabriel Mendoza Estrada, Matej Vukovic, Simone Kopeinik, Roman Kern Published: 2026-01-19 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2601.14311v1 Large language models (LLMs) are deployed at scale, yet their training data life cycle remains opaque. This survey synthesizes research from the past ten years on three tightly coupled axes: (1) data provenance, (2) transparency, and (3) traceability, and three supporting pillars: (4) bias \& uncertainty, (5) data privacy, and (6) tools and techniques that operationalize them. A central contribution is a proposed taxonomy defining the field's domains and listing corresponding artifacts. Through analysis of 95 publications, this work identifies key methodologies concerning data generation, watermarking, bias measurement, data curation, data privacy, and the inherent trade-off between transparency and opacity. --- ### CORVUS: Red-Teaming Hallucination Detectors via Internal Signal Camouflage in Large Language Models Authors: Nay Myat Min, Long H. Pham, Hongyu Zhang, Jun Sun Published: 2026-01-19 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2601.14310v1 Single-pass hallucination detectors rely on internal telemetry (e.g., uncertainty, hidden-state geometry, and attention) of large language models, implicitly assuming hallucinations leave separable traces in these signals. We study a white-box, model-side adversary that fine-tunes lightweight LoRA adapters on the model while keeping the detector fixed, and introduce CORVUS, an efficient red-teaming procedure that learns to camouflage detector-visible telemetry under teacher forcing, including an embedding-space FGSM attention stress test. Trained on 1,000 out-of-distribution Alpaca instructions (<0.5% trainable parameters), CORVUS transfers to FAVA-Annotation across Llama-2, Vicuna, Llama-3, and Qwen2.5, and degrades both training-free detectors (e.g., LLM-Check) and probe-based detectors (e.g., SEP, ICR-probe), motivating adversary-aware auditing that incorporates external grounding or cross-model evidence. --- ### Graph Neural Networks are Heuristics Authors: Yimeng Min, Carla P. Gomes Published: 2026-01-19 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2601.13465v4 Graph neural networks are usually treated as auxiliaries for combinatorial optimization: they imitate algorithms, guide search, or supply scores to classical procedures. We show that this auxiliary role is not intrinsic. A GNN can itself be a heuristic. For the Euclidean Travelling Salesman Problem, we train a non-autoregressive GNN with no labels, rewards, sequential decoding, search, or local improvement. A differentiable Hamiltonian-cycle objective is the only supervision. The trained model produces a complete tour in one forward pass, while dropout and snapshots from a single training trajectory provide solution diversity without engineered moves. The heuristic is therefore learned, not programmed. It is also fast: batched inference remains in the millisecond regime on GPUs. Experiments on TSP100, TSP200, and TSP500 show that the model consistently improves over nearest-neighbor greedy baselines. These results identify unsupervised GNNs as a class of fast learned heuristics for combinatorial optimization. --- ### Context and Transcripts Improve Detection of Deepfake Audios of Public Figures Authors: Chongyang Gao, Marco Postiglione, Julian Baldwin, Natalia Denisenko, Isabel Gortner, Luke Fosdick, Chiara Pulice, Sarit Kraus, V. S. Subrahmanian Published: 2026-01-19 Categories: cs.AI, cs.SD arXiv: https://arxiv.org/abs/2601.13464v1 Humans use context to assess the veracity of information. However, current audio deepfake detectors only analyze the audio file without considering either context or transcripts. We create and analyze a Journalist-provided Deepfake Dataset (JDD) of 255 public deepfakes which were primarily contributed by over 70 journalists since early 2024. We also generate a synthetic audio dataset (SYN) of dead public figures and propose a novel Context-based Audio Deepfake Detector (CADD) architecture. In addition, we evaluate performance on two large-scale datasets: ITW and P$^2$V. We show that sufficient context and/or the transcript can significantly improve the efficacy of audio deepfake detectors. Performance (measured via F1 score, AUC, and EER) of multiple baseline audio deepfake detectors and traditional classifiers can be improved by 5%-37.58% in F1-score, 3.77%-42.79% in AUC, and 6.17%-47.83% in EER. We additionally show that CADD, via its use of context and/or transcripts, is more robust to 5 adversarial evasion strategies, limiting performance degradation to an average of just -0.71% across all experiments. Code, models, and datasets are available at our project page: https://sites.northwestern.edu/nsail/cadd-context-based-audio-deepfake-detection (access restricted during review). --- ### Quantum Qualifiers for Neural Network Model Selection in Hadronic Physics Authors: Brandon B. Le, D. Keller Published: 2026-01-19 Categories: cs.LG, hep-ph, nucl-th, quant-ph arXiv: https://arxiv.org/abs/2601.13463v1 As quantum machine-learning architectures mature, a central challenge is no longer their construction, but identifying the regimes in which they offer practical advantages over classical approaches. In this work, we introduce a framework for addressing this question in data-driven hadronic physics problems by developing diagnostic tools - centered on a quantitative quantum qualifier - that guide model selection between classical and quantum deep neural networks based on intrinsic properties of the data. Using controlled classification and regression studies, we show how relative model performance follows systematic trends in complexity, noise, and dimensionality, and how these trends can be distilled into a predictive criterion. We then demonstrate the utility of this approach through an application to Compton form factor extraction from deeply virtual Compton scattering, where the quantum qualifier identifies kinematic regimes favorable to quantum models. Together, these results establish a principled framework for deploying quantum machine-learning tools in precision hadronic physics. --- ### SpatialBench-UC: Uncertainty-Aware Evaluation of Spatial Prompt Following in Text-to-Image Generation Authors: Amine Rostane Published: 2026-01-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.13462v1 Evaluating whether text-to-image models follow explicit spatial instructions is difficult to automate. Object detectors may miss targets or return multiple plausible detections, and simple geometric tests can become ambiguous in borderline cases. Spatial evaluation is naturally a selective prediction problem, the checker may abstain when evidence is weak and report confidence so that results can be interpreted as a risk coverage tradeoff rather than a single score. We introduce SpatialBench-UC, a small, reproducible benchmark for pairwise spatial relations. The benchmark contains 200 prompts (50 object pairs times 4 relations) grouped into 100 counterfactual pairs obtained by swapping object roles. We release a benchmark package, versioned prompts, pinned configs, per-sample checker outputs, and report tables, enabling reproducible and auditable comparisons across models. We also include a lightweight human audit used to calibrate the checker's abstention margin and confidence threshold. We evaluate three baselines, Stable Diffusion 1.5, SD 1.5 BoxDiff, and SD 1.4 GLIGEN. The checker reports pass rate and coverage as well as conditional pass rates on decided samples. The results show that grounding methods substantially improve both pass rate and coverage, while abstention remains a dominant factor due mainly to missing detections. --- ### Labels or Preferences? Budget-Constrained Learning with Human Judgments over AI-Generated Outputs Authors: Zihan Dong, Xiaotian Hou, Ruijia Wu, Linjun Zhang Published: 2026-01-19 Categories: stat.ML, cs.AI, cs.LG, math.ST arXiv: https://arxiv.org/abs/2601.13458v2 The increasing reliance on human preference feedback to judge AI-generated pseudo labels has created a pressing need for principled, budget-conscious data acquisition strategies. We address the crucial question of how to optimally allocate a fixed annotation budget between ground-truth labels and pairwise preferences in AI. Our solution, grounded in semi-parametric inference, casts the budget allocation problem as a monotone missing data framework. Building on this formulation, we introduce Preference-Calibrated Active Learning (PCAL), a novel method that learns the optimal data acquisition strategy and develops a statistically efficient estimator for functionals of the data distribution. Theoretically, we prove the asymptotic optimality of our PCAL estimator and establish a key robustness guarantee that ensures robust performance even with poorly estimated nuisance models. Our flexible framework applies to a general class of problems, by directly optimizing the estimator's variance instead of requiring a closed-form solution. This work provides a principled and statistically efficient approach for budget-constrained learning in modern AI. Simulations and real-data analysis demonstrate the practical benefits and superior performance of our proposed method. --- ### Federated Learning Under Temporal Drift -- Mitigating Catastrophic Forgetting via Experience Replay Authors: Sahasra Kokkula, Daniel David, Aaditya Baruah Published: 2026-01-19 Categories: cs.LG, cs.DC arXiv: https://arxiv.org/abs/2601.13456v1 Federated Learning struggles under temporal concept drift where client data distributions shift over time. We demonstrate that standard FedAvg suffers catastrophic forgetting under seasonal drift on Fashion-MNIST, with accuracy dropping from 74% to 28%. We propose client-side experience replay, where each client maintains a small buffer of past samples mixed with current data during local training. This simple approach requires no changes to server aggregation. Experiments show that a 50-sample-per-class buffer restores performance to 78-82%, effectively preventing forgetting. Our ablation study reveals a clear memory-accuracy trade-off as buffer size increases. --- ### PhysicsSolutionAgent: Towards Multimodal Explanations for Numerical Physics Problem Solving Authors: Aditya Thole, Anmol Agrawal, Arnav Ramamoorthy, Dhruv Kumar Published: 2026-01-19 Categories: cs.CL, cs.HC arXiv: https://arxiv.org/abs/2601.13453v1 Explaining numerical physics problems often requires more than text-based solutions; clear visual reasoning can substantially improve conceptual understanding. While large language models (LLMs) demonstrate strong performance on many physics questions in textual form, their ability to generate long, high-quality visual explanations remains insufficiently explored. In this work, we introduce PhysicsSolutionAgent (PSA), an autonomous agent that generates physics-problem explanation videos of up to six minutes using Manim animations. To evaluate the generated videos, we design an assessment pipeline that performs automated checks across 15 quantitative parameters and incorporates feedback from a vision-language model (VLM) to iteratively improve video quality. We evaluate PSA on 32 videos spanning numerical and theoretical physics problems. Our results reveal systematic differences in video quality depending on problem difficulty and whether the task is numerical or theoretical. Using GPT-5-mini, PSA achieves a 100% video-completion rate with an average automated score of 3.8/5. However, qualitative analysis and human inspection uncover both minor and major issues, including visual layout inconsistencies and errors in how visual content is interpreted during feedback. These findings expose key limitations in reliable Manim code generation and highlight broader challenges in multimodal reasoning and evaluation for visual explanations of numerical physics problems. Our work underscores the need for improved visual understanding, verification, and evaluation frameworks in future multimodal educational systems --- ### Event-based Heterogeneous Information Processing for Online Vision-based Obstacle Detection and Localization Authors: Reza Ahmadvand, Sarah Safura Sharif, Yaser Mike Banad Published: 2026-01-19 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2601.13451v1 This paper introduces a novel framework for robotic vision-based navigation that integrates Hybrid Neural Networks (HNNs) with Spiking Neural Network (SNN)-based filtering to enhance situational awareness for unmodeled obstacle detection and localization. By leveraging the complementary strengths of Artificial Neural Networks (ANNs) and SNNs, the system achieves both accurate environmental understanding and fast, energy-efficient processing. The proposed architecture employs a dual-pathway approach: an ANN component processes static spatial features at low frequency, while an SNN component handles dynamic, event-based sensor data in real time. Unlike conventional hybrid architectures that rely on domain conversion mechanisms, our system incorporates a pre-developed SNN-based filter that directly utilizes spike-encoded inputs for localization and state estimation. Detected anomalies are validated using contextual information from the ANN pathway and continuously tracked to support anticipatory navigation strategies. Simulation results demonstrate that the proposed method offers acceptable detection accuracy while maintaining computational efficiency close to SNN-only implementations, which operate at a fraction of the resource cost. This framework represents a significant advancement in neuromorphic navigation systems for robots operating in unpredictable and dynamic environments. --- ### Fairness-informed Pareto Optimization : An Efficient Bilevel Framework Authors: Sofiane Tanji, Samuel Vaiter, Yassine Laguel Published: 2026-01-19 Categories: cs.LG, math.OC, stat.ML arXiv: https://arxiv.org/abs/2601.13448v2 Despite their promise, fair machine learning methods often yield Pareto-inefficient models, in which the performance of certain groups can be improved without degrading that of others. This issue arises frequently in traditional in-processing approaches such as fairness-through-regularization. In contrast, existing Pareto-efficient approaches are biased towards a certain perspective on fairness and fail to adapt to the broad range of fairness metrics studied in the literature. In this paper, we present BADR, a simple framework to recover the optimal Pareto-efficient model for any fairness metric. Our framework recovers its models through a Bilevel Adaptive Rescalarisation procedure. The lower level is a weighted empirical risk minimization task where the weights are a convex combination of the groups, while the upper level optimizes the chosen fairness objective. We equip our framework with two novel large-scale, single-loop algorithms, BADR-GD and BADR-SGD, and establish their convergence guarantees. We release badr, an open-source Python toolbox implementing our framework for a variety of learning tasks and fairness metrics. Finally, we conduct extensive numerical experiments demonstrating the advantages of BADR over existing Pareto-efficient approaches to fairness. --- ### BladeSDF : Unconditional and Conditional Generative Modeling of Representative Blade Geometries Using Signed Distance Functions Authors: Ashish S. Nair, Sandipp Krishnan Ravi, Itzel Salgado, Changjie Sun, Sayan Ghosh, Liping Wang Published: 2026-01-19 Categories: cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2601.13445v1 Generative AI has emerged as a transformative paradigm in engineering design, enabling automated synthesis and reconstruction of complex 3D geometries while preserving feasibility and performance relevance. This paper introduces a domain-specific implicit generative framework for turbine blade geometry using DeepSDF, addressing critical gaps in performance-aware modeling and manufacturable design generation. The proposed method leverages a continuous signed distance function (SDF) representation to reconstruct and generate smooth, watertight geometries with quantified accuracy. It establishes an interpretable, near-Gaussian latent space that aligns with blade-relevant parameters, such as taper and chord ratios, enabling controlled exploration and unconditional synthesis through interpolation and Gaussian sampling. In addition, a compact neural network maps engineering descriptors, such as maximum directional strains, to latent codes, facilitating the generation of performance-informed geometry. The framework achieves high reconstruction fidelity, with surface distance errors concentrated within $1\%$ of the maximum blade dimension, and demonstrates robust generalization to unseen designs. By integrating constraints, objectives, and performance metrics, this approach advances beyond traditional 2D-guided or unconstrained 3D pipelines, offering a practical and interpretable solution for data-driven turbine blade modeling and concept generation. --- ### Explicit Cognitive Allocation: A Principle for Governed and Auditable Inference in Large Language Models Authors: Héctor Manuel Manzanilla-Granados, Zaira Navarrete-Cazales, Miriam Pescador-Rojas, Tonahtiu Ramírez-Romero Published: 2026-01-19 Categories: cs.AI arXiv: https://arxiv.org/abs/2601.13443v1 The rapid adoption of large language models (LLMs) has enabled new forms of AI-assisted reasoning across scientific, technical, and organizational domains. However, prevailing modes of LLM use remain cognitively unstructured: problem framing, knowledge exploration, retrieval, methodological awareness, and explanation are typically collapsed into a single generative process. This cognitive collapse limits traceability, weakens epistemic control, and undermines reproducibility, particularly in high-responsibility settings. We introduce Explicit Cognitive Allocation, a general principle for structuring AI-assisted inference through the explicit separation and orchestration of epistemic functions. We instantiate this principle in the Cognitive Universal Agent (CUA), an architecture that organizes inference into distinct stages of exploration and framing, epistemic anchoring, instrumental and methodological mapping, and interpretive synthesis. Central to this framework is the notion of Universal Cognitive Instruments (UCIs), which formalize heterogeneous means, including computational, experimental, organizational, regulatory, and educational instruments, through which abstract inquiries become investigable. We evaluate the effects of explicit cognitive and instrumental allocation through controlled comparisons between CUA-orchestrated inference and baseline LLM inference under matched execution conditions. Across multiple prompts in the agricultural domain, CUA inference exhibits earlier and structurally governed epistemic convergence, higher epistemic alignment under semantic expansion, and systematic exposure of the instrumental landscape of inquiry. In contrast, baseline LLM inference shows greater variability in alignment and fails to explicitly surface instrumental structure. --- ### Analyzing VLM-Based Approaches for Anomaly Classification and Segmentation Authors: Mohit Kakda, Mirudula Shri Muthukumaran, Uttapreksha Patel, Lawrence Swaminathan Xavier Prince Published: 2026-01-19 Categories: cs.CV arXiv: https://arxiv.org/abs/2601.13440v1 Vision-Language Models (VLMs), particularly CLIP, have revolutionized anomaly detection by enabling zero-shot and few-shot defect identification without extensive labeled datasets. By learning aligned representations of images and text, VLMs facilitate anomaly classification and segmentation through natural language descriptions of normal and abnormal states, eliminating traditional requirements for task-specific training or defect examples. This project presents a comprehensive analysis of VLM-based approaches for anomaly classification (AC) and anomaly segmentation (AS). We systematically investigate key architectural paradigms including sliding window-based dense feature extraction (WinCLIP), multi-stage feature alignment with learnable projections (AprilLab framework), and compositional prompt ensemble strategies. Our analysis evaluates these methods across critical dimensions: feature extraction mechanisms, text-visual alignment strategies, prompt engineering techniques, zero-shot versus few-shot trade-offs, computational efficiency, and cross-domain generalization. Through rigorous experimentation on benchmarks such as MVTec AD and VisA, we compare classification accuracy, segmentation precision, and inference efficiency. The primary contribution is a foundational understanding of how and why VLMs succeed in anomaly detection, synthesizing practical insights for method selection and identifying current limitations. This work aims to facilitate informed adoption of VLM-based methods in industrial quality control and guide future research directions. --- ### MOSLD-Bench: Multilingual Open-Set Learning and Discovery Benchmark for Text Categorization Authors: Adriana-Valentina Costache, Daria-Nicoleta Dragomir, Silviu-Florin Gheorghe, Eduard Poesina, Paul Irofti, Radu Tudor Ionescu Published: 2026-01-19 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2601.13437v1 Open-set learning and discovery (OSLD) is a challenging machine learning task in which samples from new (unknown) classes can appear at test time. It can be seen as a generalization of zero-shot learning, where the new classes are not known a priori, hence involving the active discovery of new classes. While zero-shot learning has been extensively studied in text classification, especially with the emergence of pre-trained language models, open-set learning and discovery is a comparatively new setup for the text domain. To this end, we introduce the first multilingual open-set learning and discovery (MOSLD) benchmark for text categorization by topic, comprising 960K data samples across 12 languages. To construct the benchmark, we (i) rearrange existing datasets and (ii) collect new data samples from the news domain. Moreover, we propose a novel framework for the OSLD task, which integrates multiple stages to continuously discover and learn new classes. We evaluate several language models, including our own, to obtain results that can be used as reference for future work. We release our benchmark at https://github.com/Adriana19Valentina/MOSLD-Bench. --- ### Distribution-Free Confidence Ellipsoids for Ridge Regression with PAC Bounds Authors: Szabolcs Szentpéteri, Balázs Csanád Csáji Published: 2026-01-19 Categories: stat.ML, cs.LG, eess.SP, eess.SY, math.ST arXiv: https://arxiv.org/abs/2601.13436v1 Linearly parametrized models are widely used in control and signal processing, with the least-squares (LS) estimate being the archetypical solution. When the input is insufficiently exciting, the LS problem may be unsolvable or numerically unstable. This issue can be resolved through regularization, typically with ridge regression. Although regularized estimators reduce the variance error, it remains important to quantify their estimation uncertainty. A possible approach for linear regression is to construct confidence ellipsoids with the Sign-Perturbed Sums (SPS) ellipsoidal outer approximation (EOA) algorithm. The SPS EOA builds non-asymptotic confidence ellipsoids under the assumption that the noises are independent and symmetric about zero. This paper introduces an extension of the SPS EOA algorithm to ridge regression, and derives probably approximately correct (PAC) upper bounds for the resulting region sizes. Compared with previous analyses, our result explicitly show how the regularization parameter affects the region sizes, and provide tighter bounds under weaker excitation assumptions. Finally, the practical effect of regularization is also demonstrated via simulation experiments. --- ### A Learnable Wavelet Transformer for Long-Short Equity Trading and Risk-Adjusted Return Optimization Authors: Shuozhe Li, Du Cheng, Leqi Liu Published: 2026-01-19 Categories: cs.LG, cs.AI, q-fin.CP arXiv: https://arxiv.org/abs/2601.13435v4 Learning profitable intraday trading policies from financial time series is challenging due to heavy noise, non-stationarity, and strong cross-sectional dependence among related assets. We propose \emph{WaveLSFormer}, a learnable wavelet-based long-short Transformer that jointly performs multi-scale decomposition and return-oriented decision learning. Unlike standard time-series forecasting that optimizes prediction error and typically requires a separate position-sizing or portfolio-construction step, our model directly outputs a market-neutral long/short portfolio and is trained end-to-end on a trading objective with risk-aware regularization. Specifically, a learnable wavelet front-end generates low-/high-frequency components via an end-to-end trained filter bank, guided by spectral regularizers that encourage stable and well-separated frequency bands. To fuse multi-scale information, we introduce a low-guided high-frequency injection (LGHI) module that refines low-frequency representations with high-frequency cues while controlling training stability. The model outputs a portfolio of long/short positions that is rescaled to satisfy a fixed risk budget and is optimized directly with a trading objective and risk-aware regularization. Extensive experiments on five years of hourly data across six industry groups, evaluated over ten random seeds, demonstrate that WaveLSFormer consistently outperforms MLP, LSTM and Transformer backbones, with and without fixed discrete wavelet front-ends. On average in all industries, WaveLSFormer achieves a cumulative overall strategy return of $0.607 \pm 0.045$ and a Sharpe ratio of $2.157 \pm 0.166$, substantially improving both profitability and risk-adjusted returns over the strongest baselines. ---