# Research Papers — 2026-02 · bundle page 67 of 210 > Full-text markdown of papers #3301–3350 of 10493 published in 2026-02, newest first. AI-readable mirrors of the full 10493-paper 2026-02 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-02/page/67 [← Previous bundle](https://aipod.dev/papers/archive/2026-02/page/66.md) [Next bundle →](https://aipod.dev/papers/archive/2026-02/page/68.md) ### Know When You're Wrong: Aligning Confidence with Correctness for LLM Error Detection Authors: Xie Xiaohu, Liu Xiaohu, Yao Benjamin Published: 2026-02-18 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.06604v1 As large language models (LLMs) are increasingly deployed in critical decision-making systems, the lack of reliable methods to measure their uncertainty presents a fundamental trustworthiness risk. We introduce a normalized confidence score based on output anchor token probabilities: classification labels for structured tasks and self-evaluation responses (Yes/No) for open-ended generation. This enables direct detection of errors and hallucinations with minimal overhead and without external validation. We make three key contributions. First, we propose a normalized confidence score and self-evaluation framework that exposes reliable confidence estimates for error detection across seven diverse benchmark tasks and five LLMs of varying architectures and sizes. Second, our theoretical analysis reveals that supervised fine-tuning (SFT) yields well-calibrated confidence through maximum-likelihood estimation, whereas reinforcement learning methods (PPO, GRPO) and DPO induce overconfidence via reward exploitation. Third, we propose post-RL SFT with self-distillation to restore confidence reliability in RL-trained models. Empirical results demonstrated that SFT improved average confidence-correctness AUROC from 0.806 to 0.879 and reduced calibration error from 0.163 to 0.034 on Qwen3-4B, while GRPO and DPO degraded confidence reliability. We demonstrated practical value through adaptive retrieval-augmented generation (RAG) that selectively retrieves context when the model lacks confidence, using only 58\% of retrieval operations to recover 95\% of the maximum achievable accuracy gain on TriviaQA --- ### Scale Dependent Data Duplication Authors: Joshua Kazdan, Noam Levi, Rylan Schaeffer, Jessica Chudnovsky, Abhay Puri, Bo He, Mehmet Donmez, Sanmi Koyejo, David Donoho Published: 2026-02-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.06603v1 Data duplication during pretraining can degrade generalization and lead to memorization, motivating aggressive deduplication pipelines. However, at web scale, it is unclear what constitutes a ``duplicate'': beyond surface-form matches, semantically equivalent documents (e.g. translations) may induce redundant training signals once models become sufficiently capable. Practically, this means that semantic duplicates operate increasingly like exact duplicates during training. We present evidence that duplication is scale-dependent in two ways. First, as model capability increases, cross-entropy loss gradients for semantically equivalent documents become more aligned. Smaller models, by contrast, produce gradients that reflect surface similarity (e.g., shared tokens) rather than semantic similarity. Second, we embedded all 192 million FineWeb-Edu-Dedup documents using EmbeddingGemma-300m. For moderate corpus sizes, the cosine similarity between nearest-neighbors follows an isotropic power law baseline. However, as corpus size grows to hundreds of billions of tokens, the nearest-neighbor similarities deviate sharply, indicating accelerated semantic collisions. Finally, controlled pretraining on data sampled with replacement from pools of finite unique documents shows that limited uniqueness yields mild degradation for small models, but rapidly increasing loss penalties for larger models, breaking naive scaling extrapolation. We derive explicit scaling laws that allow practitioners to estimate deviation from expected scaling due to limited semantic uniqueness of the pretraining corpus. Our results identify and resolve an unstudied source of scale-dependence, allowing for more accurate prediction at scale. --- ### RoboLayout: Differentiable 3D Scene Generation for Embodied Agents Authors: Ali Shamsaddinlou Published: 2026-02-18 Categories: cs.AI, cs.CV, cs.LG, cs.RO arXiv: https://arxiv.org/abs/2603.05522v2 Recent advances in vision language models (VLMs) have shown strong potential for spatial reasoning and 3D scene layout generation from open-ended language instructions. However, generating layouts that are not only semantically coherent but also feasible for interaction by embodied agents remains challenging, particularly in physically constrained indoor environments. In this paper, RoboLayout is introduced as an extension of LayoutVLM that augments the original framework with agent-aware reasoning and improved optimization stability. RoboLayout integrates explicit reachability constraints into a differentiable layout optimization process, enabling the generation of layouts that are navigable and actionable by embodied agents. Importantly, the agent abstraction is not limited to a specific robot platform and can represent diverse entities with distinct physical capabilities, such as service robots, warehouse robots, humans of different age groups, or animals, allowing environment design to be tailored to the intended agent. In addition, a local refinement stage is proposed that selectively reoptimizes problematic object placements while keeping the remainder of the scene fixed, improving convergence efficiency without increasing global optimization iterations. Overall, RoboLayout preserves the strong semantic alignment and physical plausibility of LayoutVLM while enhancing applicability to agent-centric indoor scene generation, as demonstrated by experimental results across diverse scene configurations. --- ### Auction-Based RIS Allocation With DRL: Controlling the Cost-Performance Trade-Off Authors: Martin Mark Zan, Stefan Schwarz Published: 2026-02-18 Categories: cs.NI, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2603.04433v1 We study the allocation of reconfigurable intelligent surfaces (RISs) in a multi-cell wireless network, where base stations compete for control of shared RIS units deployed at the cell edges. These RISs, provided by an independent operator, are dynamically leased to the highest bidder using a simultaneously ascending auction format. Each base station estimates the utility of acquiring additional RISs based on macroscopic channel parameters, enabling a scalable and low-overhead allocation mechanism. To optimize the bidding behavior, we integrate deep reinforcement learning (DRL) agents that learn to maximize performance while adhering to budget constraints. Through simulations in clustered cell-edge environments, we demonstrate that reinforcement learning (RL)-based bidding significantly outperforms heuristic strategies, achieving optimal trade-offs between cost and spectral efficiency. Furthermore, we introduce a tunable parameter that governs the bidding aggressiveness of RL agents, enabling a flexible control of the trade-off between network performance and expenditure. Our results highlight the potential of combining auction-based allocation with adaptive RL mechanisms for efficient and fair utilization of RISs in next-generation wireless networks. --- ### Cryo-SWAN: the Multi-Scale Wavelet-decomposition-inspired Autoencoder Network for molecular density representation of molecular volumes Authors: Rui Li, Artsemi Yushkevich, Mikhail Kudryashev, Artur Yakimovich Published: 2026-02-18 Categories: eess.IV, cs.AI, q-bio.BM arXiv: https://arxiv.org/abs/2603.03342v1 Learning robust representations of 3D shapes from voxelized data is essential for advancing AI methods in biomedical imaging. However, most contemporary 3D computer vision approaches operate on point clouds, meshes, or octrees, while volumetric density maps, the native format of structural biology and cryo-EM, remain comparatively underexplored. We present Cryo-SWAN, a voxel-based variational autoencoder inspired by multi-scale wavelet decomposition. The model performs conditional coarse-to-fine latent encoding and recursive residual quantization across perception scales, enabling accurate capture of both global geometry and high-frequency structural detail in molecular density volumes. Evaluated on ModelNet40, BuildingNet, and a newly curated dataset of cryo-EM volumes, ProteinNet3D, Cryo-SWAN consistently improves reconstruction quality over state-of-the-art 3D autoencoders. We demonstrate that the molecular densities organize in learned latent space according to shared geometric features, while integration with diffusion models enables denoising and conditional shape generation. Together, Cryo-SWAN is a practical framework for data-driven structural biology and volumetric imaging. --- ### RSS map-assisted MIMO channel estimation in the upper mid-band under pilot constraints Authors: Alireza Javid, Nuria González-Prelcic Published: 2026-02-18 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2603.00112v1 Accurate wireless channel estimation is critical for next-generation wireless systems, enabling precise precoding for effective user separation, reduced interference across cells, and high-resolution sensing, among other benefits. Traditional model-based channel estimation methods suffer, however, from performance degradation in complex environments with a limited number of pilots, while purely data-driven approaches lack physical interpretability, require extensive data collection, and are usually site-specific. This paper presents a novel physics-informed neural network (PINN) framework that synergistically combines model-based channel estimation with a deep network to exploit prior information about environmental propagation characteristics and achieve superior performance under pilot-constrained scenarios. The proposed approach employs an enhanced U-Net architecture with transformer modules and cross-attention mechanisms to fuse initial channel estimates with RSS maps to provide refined channel estimates. Comprehensive evaluation using realistic ray-tracing data from urban environments demonstrates significant performance improvements, achieving over 5 dB gain in NMSE compared to state-of-the-art methods, with particularly strong performance in pilot-limited scenarios and robustness across different frequencies and environments with only minimal fine-tuning. We further extend the decoder for multi-step temporal prediction, enabling accurate forecasting of several future channel snapshots from a single estimate, useful for proactive beamforming and scheduling in mobile scenarios. The proposed framework maintains practical computational complexity, making it viable for massive MIMO systems in upper-mid band frequencies. Unlike black-box neural approaches, the physics-informed design provides a more interpretable channel estimation method. --- ### CASCADE: Cross-scale Advective Super-resolution with Climate Assimilation and Downscaling Evolution Authors: Alexander Kovalenko Published: 2026-02-18 Categories: physics.ao-ph, cs.LG arXiv: https://arxiv.org/abs/2603.00109v1 Super-resolution of geophysical fields presents unique challenges beyond natural image enhancement: fine-scale structures must respect physical dynamics, conserve mass and energy, and evolve coherently in time. These constraints are especially critical for extreme events, where rare, localized, high-intensity features drive impacts and where temporally inconsistent "hallucinated" detail can misrepresent hazards. We introduce CASCADE (Cross-scale Advective Super-resolution with Climate Assimilation and Downscaling Evolution), a framework that reframes spatiotemporal super-resolution as an explicit transport process across scales. Rather than hallucinating high-frequency content per pixel, CASCADE reconstructs fine structure by iteratively advecting coarse information along learned, flow-conditioned velocity fields through semi-Lagrangian warping. The architecture decomposes motion into resolved (large-scale) and subgrid (unresolved) components, mirroring the closure problem in numerical weather prediction, and enforces low-resolution consistency through an assimilation-style innovation step. Evaluated on SEVIR radar data for 4x super-resolution of severe convective storms, CASCADE outperforms strong baselines across both continuous metrics (PSNR, SSIM, MAE) and threshold-based skill scores (CSI, HSS, POD) while providing interpretable diagnostics through visualizable velocity and correction fields. By encoding advection as the fundamental operator rather than learning it implicitly, CASCADE produces temporally coherent, physically consistent, and mass-conserving reconstructions well suited to advection-dominated extremes in atmospheric and oceanic applications. --- ### SurgFusion-Net: Diversified Adaptive Multimodal Fusion Network for Surgical Skill Assessment Authors: Runlong He, Freweini M. Tesfai, Matthew W. E. Boal, Nazir Sirajudeen, Dimitrios Anastasiou, Jialang Xu, Mobarak I. Hoque, Philip J. Edwards, John D. Kelly, Ashwin Sridhar, Abdolrahim Kadkhodamohammadi, Dhivya Chandrasekaran et al. Published: 2026-02-18 Categories: cs.RO, cs.AI arXiv: https://arxiv.org/abs/2603.00108v1 Robotic-assisted surgery (RAS) is established in clinical practice, and automated surgical skill assessment utilizing multimodal data offers transformative potential for surgical analytics and education. However, developing effective multimodal methods remains challenging due to the task complexity, limited annotated datasets and insufficient techniques for cross-modal information fusion. Existing state-of-the-art relies exclusively on RGB video and only applies on dry-lab settings, failing to address the significant domain gap between controlled simulation and real clinical cases, where the surgical environment together with camera and tissue motion introduce substantial complexities. This work introduces SurgFusion-Net and Divergence Regulated Attention (DRA), an innovative fusion strategy for multimodal surgical skill assessment. We contribute two first-of-their-kind clinical datasets: the RAH-skill dataset containing 279,691 RGB frames from 37 videos of Robot-assisted Hysterectomy (RAH), and the RARP-skill dataset containing 70,661 RGB frames from 33 videos of Robot-Assisted Radical Prostatectomy (RARP). Both datasets include M-GEARS skill annotations, corresponding optical flow and tool segmentation masks. DRA incorporates adaptive dual attention and diversity-promoting multi-head attention to fuse multimodal information, from three modalities, based on surgical context, enhancing assessment accuracy and reliability. Validated on the JIGSAWS benchmark, RAH-skill, and RARP-skill datasets, our approach outperforms recent baselines with SCC improvements of 0.02 in LOSO, 0.04 in LOUO across JIGSAWS tasks, and 0.0538 and 0.0493 gains on RAH-skill and RARP-skill, respectively. --- ### LIDS: LLM Summary Inference Under the Layered Lens Authors: Dylan Park, Yingying Fan, Jinchi Lv Published: 2026-02-18 Categories: cs.LG, cs.CL, stat.ME, stat.ML arXiv: https://arxiv.org/abs/2603.00105v1 Large language models (LLMs) have gained significant attention by many researchers and practitioners in natural language processing (NLP) since the introduction of ChatGPT in 2022. One notable feature of ChatGPT is its ability to generate summaries based on prompts. Yet evaluating the quality of these summaries remains challenging due to the complexity of language. To this end, in this paper we suggest a new method of LLM summary inference with BERT-SVD-based direction metric and SOFARI (LIDS) that assesses the summary accuracy equipped with interpretable key words for layered themes. The LIDS uses a latent SVD-based direction metric to measure the similarity between the summaries and original text, leveraging the BERT embeddings and repeated prompts to quantify the statistical uncertainty. As a result, LIDS gives a natural embedding of each summary for large text reduction. We further exploit SOFARI to uncover important key words associated with each latent theme in the summary with controlled false discovery rate (FDR). Comprehensive empirical studies demonstrate the practical utility and robustness of LIDS through human verification and comparisons to other similarity metrics, including a comparison of different LLMs. --- ### Alpha-RF: Automated RF-Filter-Circuit Design with Neural Simulator and Reinforcement Learning Authors: Nhat Tran, Chenjie Hao, Alexander Stameroff, Anh-Vu Pham, Yubei Chen Published: 2026-02-18 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2603.00104v1 Accurate, high-performance radio-frequency (RF) filter circuits are ubiquitous in radio-frequency communication and sensing systems for accepting and rejecting signals at desired frequencies. Conventional RF filter design process involves manual calculations of design parameters, followed by intuition-guided iterations to achieve the desired response for a set of filter specifications. This process is time-consuming due to time- and resource-intensive electromagnetic simulations using full-wave numerical PDE solvers. This process is also highly sensitive to domain expertise and requires many years of professional training. To address these bottlenecks, we propose an automatic RF filter circuit design tool using neural simulator and reinforcement learning. First, we train a neural simulator to replace the PDE electromagnetic simulator. The neural-network-based simulator reduces each of the simulation time from 4 minutes on average to less than 100 millisecond while maintaining a high precision. Such dramatic acceleration enable us to leverage deep reinforcement learning algorithm and train an amortized inference policy to perform automatic design in the imagined space from the neural simulator. The resulted automatic circuit-design agent achieves super-human design results. The automatic circuit-design agent also reduces the on-average design cycle from days to under a few seconds. Even more surprisingly, we demonstrate that the neural simulator can generalize to design spaces far from the training dataset and in a sense it has learned the underlying physics--Maxwell equations. We also demonstrate that the reinforcement learning has discovered many expert-like design intuitions. This work marks a step in using neural simulators and reinforcement learning in RF circuit design and the proposed method is generally applicable to many other design problems and domains in close affinity --- ### Depth from Defocus via Direct Optimization Authors: Holly Jackson, Caleb Adams, Ignacio Lopez-Francos, Benjamin Recht Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.18509v2 Though there exists a reasonable forward model for blur based on optical physics, recovering depth from a collection of defocused images remains a computationally challenging optimization problem. In this paper, we show that with contemporary optimization methods and reasonable computing resources, a global optimization approach to depth from defocus is feasible. Our approach rests on alternating minimization. When holding the depth map fixed, the forward model is linear with respect to the all-in-focus image. When holding the all-in-focus image fixed, the depth at each pixel can be computed independently, enabling embarrassingly parallel computation. We show that alternating between convex optimization and parallel grid search can effectively solve the depth-from-defocus problem at higher resolutions than current deep learning methods. We demonstrate our approach on benchmark datasets with synthetic and real defocus blur and show promising results compared to prior approaches. Our code is available at github.com/hollyjackson/dfd. --- ### Parallelizable Neural Turing Machines Authors: Gabriel Faria, Arnaldo Candido Junior Published: 2026-02-18 Categories: cs.NE arXiv: https://arxiv.org/abs/2602.18508v1 We introduce a parallelizable simplification of Neural Turing Machine (NTM), referred to as P-NTM, which redesigns the core operations of the original architecture to enable efficient scan-based parallel execution. We evaluate the proposed architecture on a synthetic benchmark of algorithmic problems involving state tracking, memorization, and basic arithmetic, solved via autoregressive decoding. We compare it against a revisited stable implementation of the standard NTM, as well as conventional recurrent and attention-based architectures. Results show that, despite its simplifications, the proposed model attains length generalization performance comparable to the original, learning to solve all problems, including unseen sequence lengths, with perfect accuracy. It also improves training efficiency, with parallel execution of P-NTM being up to an order of magnitude faster than the standard NTM. Ultimately, this work contributes toward the development of efficient neural architectures capable of expressing a broad class of algorithms. --- ### Fine-Pruning: A Biologically Inspired Algorithm for Personalization of Machine Learning Models Authors: Joseph Bingham, Saman Zonouz, Dvir Aran Published: 2026-02-18 Categories: cs.NE, q-bio.NC arXiv: https://arxiv.org/abs/2602.18507v1 Neural networks have long strived to emulate the learning capabilities of the human brain. While deep neural networks (DNNs) draw inspiration from the brain in neuron design, their training methods diverge from biological foundations. Backpropagation, the primary training method for DNNs, requires substantial computational resources and fully labeled datasets, presenting major bottlenecks in development and application. This work demonstrates that by returning to biomimicry, specifically mimicking how the brain learns through pruning, we can solve various classical machine learning problems while utilizing orders of magnitude fewer computational resources and no labels. Our experiments successfully personalized multiple speech recognition and image classification models, including ResNet50 on ImageNet, resulting in increased sparsity of approximately 70\% while simultaneously improving model accuracy to around 90\%, all without the limitations of backpropagation. This biologically inspired approach offers a promising avenue for efficient, personalized machine learning models in resource-constrained environments. --- ### Suppression or Deletion: A Restoration-Based Representation-Level Analysis of Machine Unlearning Authors: Yurim Jang, Jaeung Lee, Dohyun Kim, Jaemin Jo, Simon S. Woo Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.18505v1 As pretrained models are increasingly shared on the web, ensuring that models can forget or delete sensitive, copyrighted, or private information upon request has become crucial. Machine unlearning has been proposed to address this challenge. However, current evaluations for unlearning methods rely on output-based metrics, which cannot verify whether information is completely deleted or merely suppressed at the representation level, where suppression is insufficient for true unlearning. To address this gap, we propose a novel restoration-based analysis framework that uses Sparse Autoencoders to identify class-specific expert features in intermediate layers and applies inference-time steering to quantitatively distinguish between suppression and deletion. Applying our framework to 12 major unlearning methods in image classification tasks, we find that most methods achieve high restoration rates of unlearned information, indicating that they only suppress information at the decision-boundary level, while preserving semantic features in intermediate representations. Notably, even retraining from pretrained checkpoints shows high restoration, revealing that robust semantic features inherited from pretraining are not removed by retraining. These results demonstrate that representation-level retention poses significant risks overlooked by output-based metrics, highlighting the need for new unlearning evaluation criteria. We propose new evaluation guidelines that prioritize representation-level verification, especially for privacy-critical applications in the era of pre-trained models. --- ### NestRL: A Nested Training Regime for Mutual Adaptation in Human-AI Teaming Authors: Upasana Biswas, Durgesh Kalwar, Subbarao Kambhampati, Sarath Sreedharan Published: 2026-02-18 Categories: cs.RO, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2602.17737v2 Mutual adaptation is a central challenge in human-AI teaming, as humans naturally adjust their strategies in response to an AI agent's behavior. Existing approaches attempt to approximate human behavior by diversifying training partners; however, these partners are typically static and fail to capture the adaptive nature of human teammates. When agents are trained jointly in standard multi-agent settings, they often converge to opaque coordination strategies that work only with their co-trained partners, leading to poor generalization. To model adaptive human behavior, we formulate human-AI teaming as an Interactive Partially Observable Markov Decision Process (I-POMDP). We propose NestRL, a nested training regime that learns the solution to a finite-level I-POMDP by training agents at each level against adaptive agents from the level below. This exposes agents to adaptive behavior while preventing emergence of opaque coordination strategies. We provide theoretical analysis showing that NestRL agents avoid convergence to partner-specific strategies, and validate this empirically in the Overcooked domain against state-of-the-art baselines. NestRL achieves higher task performance with both unseen adaptive agents and real human teammates, while exhibiting significantly greater adaptability over the course of interaction. --- ### Five Fatal Assumptions: Why T-Shirt Sizing Systematically Fails for AI Projects Authors: Raja Soundaramourty, Ozkan Kilic, Ramu Chenchaiah Published: 2026-02-18 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2602.17734v1 Agile estimation techniques, particularly T-shirt sizing, are widely used in software development for their simplicity and utility in scoping work. However, when we apply these methods to artificial intelligence initiatives -- especially those involving large language models (LLMs) and multi-agent systems -- the results can be systematically misleading. This paper shares an evidence-backed analysis of five foundational assumptions we often make during T-shirt sizing. While these assumptions usually hold true for traditional software, they tend to fail in AI contexts: (1) linear effort scaling, (2) repeatability from prior experience, (3) effort-duration fungibility, (4) task decomposability, and (5) deterministic completion criteria. Drawing on recent research into multi-agent system failures, scaling principles, and the inherent unreliability of multi-turn conversations, we show how AI development breaks these rules. We see this through non-linear performance jumps, complex interaction surfaces, and "tight coupling" where a small change in data cascades through the entire stack. To help teams navigate this, we propose Checkpoint Sizing: a more human-centric, iterative approach that uses explicit decision gates where scope and feasibility are reassessed based on what we learn during development, rather than what we assumed at the start. This paper is intended for engineering managers, technical leads, and product owners responsible for planning and delivering AI initiatives. --- ### Clever Materials: When Models Identify Good Materials for the Wrong Reasons Authors: Kevin Maik Jablonka Published: 2026-02-18 Categories: physics.chem-ph, cond-mat.mtrl-sci, cs.LG arXiv: https://arxiv.org/abs/2602.17730v1 Machine learning can accelerate materials discovery. Models perform impressively on many benchmarks. However, strong benchmark performance does not imply that a model learned chemistry. I test a concrete alternative hypothesis: that property prediction can be driven by bibliographic confounding. Across five tasks spanning MOFs (thermal and solvent stability), perovskite solar cells (efficiency), batteries (capacity), and TADF emitters (emission wavelength), models trained on standard chemical descriptors predict author, journal, and publication year well above chance. When these predicted metadata ("bibliographic fingerprints") are used as the sole input to a second model, performance is sometimes competitive with conventional descriptor-based predictors. These results show that many datasets do not rule out non-chemical explanations of success. Progress requires routine falsification tests (e.g., group/time splits and metadata ablations), datasets designed to resist spurious correlations, and explicit separation of two goals: predictive utility versus evidence of chemical understanding. --- ### Stop Saying "AI" Authors: Nathan G. Wood, Scott Robbins, Eduardo Zegarra Berodt, Anton Graf von Westerholt, Michelle Behrndt, Hauke Budig, Daniel Kloock-Schreiber Published: 2026-02-18 Categories: cs.CY, cs.AI, cs.ET, cs.HC arXiv: https://arxiv.org/abs/2602.17729v2 Across academia, industry, and government, ``AI'' has become central in research and development, regulatory debates, and promises of ever faster and more capable decision-making and action. In numerous domains, especially safety-critical ones, there are significant concerns over how ``AI'' may affect decision-making, responsibility, or the likelihood of mistakes (to name only a few categories of critique). However, for most critiques, the target is generally ``AI'', a broad term admitting many (types of) systems used for a variety of tasks and each coming with its own set of limitations, challenges, and potential use cases. In this article, we focus on the military domain as a case study and present both a loose enumerative taxonomy of systems captured under the umbrella term ``military AI'', as well as discussion of the challenges of each. In doing so, we highlight that critiques of one (type of) system will not always transfer to other (types of) systems. Building on this, we argue that in order for debates to move forward fruitfully, it is imperative that the discussions be made more precise and that ``AI'' be excised from debates to the extent possible. Researchers, developers, and policy-makers should make clear exactly what systems they have in mind and what possible benefits and risks attend the deployment of those particular systems. While we focus on AI in the military as an exemplar for the overall trends in discussions of ``AI'', the argument's conclusions are broad and have import for discussions of AI across a host of domains. --- ### Multi-Agent Lipschitz Bandits Authors: Sourav Chakraborty, Amit Kiran Rege, Claire Monteleoni, Lijun Chen Published: 2026-02-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.16965v2 We study the decentralized multi-player stochastic bandit problem over a continuous, Lipschitz-structured action space where hard collisions yield zero reward. Our objective is to design a communication-free policy that maximizes collective reward, while separating coordination costs from learning costs. We propose a modular protocol that first solves the multi-agent coordination problem by identifying and seating players on distinct, high-value regions via a novel maxima-directed search and then decouples the problem into $N$ independent single-player Lipschitz bandits. In the consensus regime, we obtain an end-to-end regret bound whose dominant learning term is \(\tilde{O}(T^{(d+1)/(d+2)})\), matching the single-player Lipschitz rate; the upfront coordination cost is horizon-independent at fixed confidence and only polylogarithmic in \(T\) in the expected-regret form. Under an additional public coverage/scheduling assumption for the epochic extension, we also obtain a gap-free \(\tilde{O}(T^{(d+1)/(d+2)})\) guarantee. We further derive a matching lower bound for the dominant learning term and extend the framework to general distance-threshold collision models. --- ### Greedy Multi-Path Block Verification for Faster Decoding in Speculative Sampling Authors: Rahul Thomas, Arka Pal Published: 2026-02-18 Categories: cs.IT, cs.LG arXiv: https://arxiv.org/abs/2602.16961v1 The goal of $L$-step speculative decoding is to accelerate autoregressive decoding of a target model by using a cheaper draft model to generate a candidate path of $L$ tokens. Based on a verification algorithm involving target and draft model probabilities, a prefix of the candidate sequence is accepted, and an additional correction token is sampled from a residual distribution to ensure that the final output adheres to the target distribution. While standard speculative decoding uses a verification algorithm which is independent at each token on the path, a recent extension called block verification uses a joint condition involving all sampled on-path probabilities. Block verification (BV) was shown to be optimal over all verification algorithms which use only on-path probabilities, improving on standard speculative decoding. In this work, we first show that block verification is optimal even over verification algorithms that use off-path probabilities, by constructing an information-agnostic linear program (LP). Further, we can extend our LP to the setting where the draft model samples multiple candidate paths, and use it to construct a natural class of multi-path block verification generalizations. While computing the optimal algorithm in this class is not tractable, by considering a stricter class of greedy algorithms, we can formulate an efficient method called greedy multi-path block verification (GBV). Empirically, GBV can improve block efficiency by over 30% and reduce decoding walltimes by over 15% relative to BV. On Llama-3 70B, GBV can improve the end-to-end decoding throughput over SOTA multi-path verification methods by more than 15%. --- ### Eigenmood Space: Uncertainty-Aware Spectral Graph Analysis of Psychological Patterns in Classical Persian Poetry Authors: Kourosh Shahnazari, Seyed Moein Ayyoubzadeh, Mohammadali Keshtparvar Published: 2026-02-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.16959v1 Classical Persian poetry is a historically sustained archive in which affective life is expressed through metaphor, intertextual convention, and rhetorical indirection. These properties make close reading indispensable while limiting reproducible comparison at scale. We present an uncertainty-aware computational framework for poet-level psychological analysis based on large-scale automatic multi-label annotation. Each verse is associated with a set of psychological concepts, per-label confidence scores, and an abstention flag that signals insufficient evidence. We aggregate confidence-weighted evidence into a Poet $\times$ Concept matrix, interpret each poet as a probability distribution over concepts, and quantify poetic individuality as divergence from a corpus baseline using Jensen--Shannon divergence and Kullback--Leibler divergence. To capture relational structure beyond marginals, we build a confidence-weighted co-occurrence graph over concepts and define an Eigenmood embedding through Laplacian spectral decomposition. On a corpus of 61{,}573 verses across 10 poets, 22.2\% of verses are abstained, underscoring the analytical importance of uncertainty. We further report sensitivity analysis under confidence thresholding, selection-bias diagnostics that treat abstention as a category, and a distant-to-close workflow that retrieves verse-level exemplars along Eigenmood axes. The resulting framework supports scalable, auditable digital-humanities analysis while preserving interpretive caution by propagating uncertainty from verse-level evidence to poet-level inference. --- ### Automating Agent Hijacking via Structural Template Injection Authors: Xinhao Deng, Jiaqing Wu, Miao Chen, Yue Xiao, Ke Xu, Qi Li Published: 2026-02-18 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.16958v1 Agent hijacking, highlighted by OWASP as a critical threat to the Large Language Model (LLM) ecosystem, enables adversaries to manipulate execution by injecting malicious instructions into retrieved content. Most existing attacks rely on manually crafted, semantics-driven prompt manipulation, which often yields low attack success rates and limited transferability to closed-source commercial models. In this paper, we propose Phantom, an automated agent hijacking framework built upon Structured Template Injection that targets the fundamental architectural mechanisms of LLM agents. Our key insight is that agents rely on specific chat template tokens to separate system, user, assistant, and tool instructions. By injecting optimized structured templates into the retrieved context, we induce role confusion and cause the agent to misinterpret the injected content as legitimate user instructions or prior tool outputs. To enhance attack transferability against black-box agents, Phantom introduces a novel attack template search framework. We first perform multi-level template augmentation to increase structural diversity and then train a Template Autoencoder (TAE) to embed discrete templates into a continuous, searchable latent space. Subsequently, we apply Bayesian optimization to efficiently identify optimal adversarial vectors that are decoded into high-potency structured templates. Extensive experiments on Qwen, GPT, and Gemini demonstrate that our framework significantly outperforms existing baselines in both Attack Success Rate (ASR) and query efficiency. Moreover, we identified over 70 vulnerabilities in real-world commercial products that have been confirmed by vendors, underscoring the practical severity of structured template-based hijacking and providing an empirical foundation for securing next-generation agentic systems. --- ### When Semantic Overlap Is Not Enough: Cross-Lingual Euphemism Transfer Between Turkish and English Authors: Hasan Can Biyik, Libby Barak, Jing Peng, Anna Feldman Published: 2026-02-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2602.16957v1 Euphemisms substitute socially sensitive expressions, often softening or reframing meaning, and their reliance on cultural and pragmatic context complicates modeling across languages. In this study, we investigate how cross-lingual equivalence influences transfer in multilingual euphemism detection. We categorize Potentially Euphemistic Terms (PETs) in Turkish and English into Overlapping (OPETs) and Non-Overlapping (NOPETs) subsets based on their functional, pragmatic, and semantic alignment. Our findings reveal a transfer asymmetry: semantic overlap is insufficient to guarantee positive transfer, particularly in low-resource Turkish-to-English direction, where performance can degrade even for overlapping euphemisms, and in some cases, improve under NOPET-based training. Differences in label distribution help explain these counterintuitive results. Category-level analysis suggests that transfer may be influenced by domain-specific alignment, though evidence is limited by sparsity. --- ### Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generative Modeling Authors: Chuqin Geng, Li Zhang, Mark Zhang, Zhaoyue Wang, Haolin Ye, Xujie Si Published: 2026-02-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.16954v3 While deep generative models excel at capturing graph data distributions, they struggle to satisfy complex, hard constraints. In unconstrained settings, these models typically produce valid topologies; yet imposing strict compositional rules, like those in drug discovery, creates an out-of-distribution (OOD) setting where purely neural methods frequently fail. Because these neural approaches rely on soft conditioning and post-hoc filtering on such tasks, they cannot provide the formal guarantees needed for high-stakes domains. To address this, we introduce Neuro-Symbolic Graph Generative Modeling (NSGGM), a framework built on the principle of Neural Proposals, Symbolic Guarantees. NSGGM decouples generation: an autoregressive model proposes structural scaffolds, and a Satisfiability Modulo Theories (SMT) solver handles the final discrete assembly of the proposed substructures. Empirically, NSGGM is competitive with state-of-the-art methods on unconstrained tasks. To evaluate logical-constraint satisfaction inspired by real drug discovery workflows, we introduce MolSAT, a benchmark for hard compositional rules. On MolSAT, purely neural baselines completely fail OOD (0% satisfaction with zero training support), while NSGGM achieves >95% satisfaction in-distribution and 64-86% with zero training support. --- ### Neural Proposals, Symbolic Guarantees: Neuro-Symbolic Graph Generation with Hard Constraints Authors: Chuqin Geng, Li Zhang, Mark Zhang, Haolin Ye, Ziyu Zhao, Xujie Si Published: 2026-02-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.16954v2 We challenge black-box purely deep neural approaches for molecules and graph generation, which are limited in controllability and lack formal guarantees. We introduce Neuro-Symbolic Graph Generative Modeling (NSGGM), a neurosymbolic framework that reapproaches molecule generation as a scaffold and interaction learning task with symbolic assembly. An autoregressive neural model proposes scaffolds and refines interaction signals, and a CPU-efficient SMT solver constructs full graphs while enforcing chemical validity, structural rules, and user-specific constraints, yielding molecules that are correct by construction and interpretable control that pure neural methods cannot provide. NSGGM delivers strong performance on both unconstrained generation and constrained generation tasks, demonstrating that neuro-symbolic modeling can match state-of-the-art generative performance while offering explicit controllability and guarantees. To evaluate more nuanced controllability, we also introduce a Logical-Constraint Molecular Benchmark, designed to test strict hard-rule satisfaction in workflows that require explicit, interpretable specifications together with verifiable compliance. --- ### LLM4Cov: Execution-Aware Agentic Learning for High-coverage Testbench Generation Authors: Hejia Zhang, Zhongming Yu, Chia-Tung Ho, Haoxing Ren, Brucek Khailany, Jishen Zhao Published: 2026-02-18 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.16953v3 Execution-aware LLM agents offer a promising paradigm for learning from tool feedback, but such feedback can be expensive and slow to obtain, making online reinforcement learning (RL) less practical in certain scenarios. High-coverage hardware verification exemplifies this challenge due to its reliance on industrial simulators and non-differentiable execution signals. We propose LLM4Cov, an offline agent-learning framework that models verification as single-step state transitions guided by deterministic evaluators. Building on this formulation, we introduce execution-validated data curation, policy-aware agentic data synthesis, and worst-state-prioritized sampling to enable scalable learning under execution constraints. We further curate a reality-aligned benchmark adapted from an existing verification suite through a revised evaluation protocol. Using the proposed pipeline, a compact 4B-parameter model achieves 69.2% pass rate and 90.4% average coverage in CVDP-ECov under agentic evaluation, outperforming its teacher by 5.3% and 10.5%, demonstrating competitive performance against models an order of magnitude larger. --- ### BrainRVQ: A High-Fidelity EEG Foundation Model via Dual-Domain Residual Quantization and Hierarchical Autoregression Authors: Mingzhe Cui, Tao Chen, Yang Jiao, Yiqin Wang, Lei Xie, Yi Pan, Luca Mainardi Published: 2026-02-18 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2602.16951v1 Developing foundation models for electroencephalography (EEG) remains challenging due to the signal's low signal-to-noise ratio and complex spectro-temporal non-stationarity. Existing approaches often overlook the hierarchical latent structure inherent in neural dynamics, leading to suboptimal reconstruction of fine-grained information. In this work, we propose BrainRVQ, a general-purpose EEG foundation model pre-trained on a large-scale corpus of clinical EEG data. Unlike standard masked modeling, BrainRVQ features a Dual-Domain Residual Vector Quantization (DD-RVQ) tokenizer that disentangles temporal waveforms and spectral patterns into hierarchical discrete codes. We further introduce a hierarchical autoregressive pre-training objective that learns to reconstruct these codes in a coarse-to-fine manner, utilizing an importance-guided curriculum masking strategy to prioritize information-rich neural events over background noise. Extensive experiments across 8 diverse downstream datasets demonstrate that BrainRVQ consistently outperforms state-of-the-art baselines, validating its effectiveness in learning robust and generalizable neural representations. Our code and model weights are available:https://github.com/keqicmz/BrainRVQ --- ### HS-3D-NeRF: 3D Surface and Hyperspectral Reconstruction From Stationary Hyperspectral Images Using Multi-Channel NeRFs Authors: Kibon Ku, Talukder Z. Jubery, Adarsh Krishnamurthy, Baskar Ganapathysubramanian Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.16950v1 Advances in hyperspectral imaging (HSI) and 3D reconstruction have enabled accurate, high-throughput characterization of agricultural produce quality and plant phenotypes, both essential for advancing agricultural sustainability and breeding programs. HSI captures detailed biochemical features of produce, while 3D geometric data substantially improves morphological analysis. However, integrating these two modalities at scale remains challenging, as conventional approaches involve complex hardware setups incompatible with automated phenotyping systems. Recent advances in neural radiance fields (NeRF) offer computationally efficient 3D reconstruction but typically require moving-camera setups, limiting throughput and reproducibility in standard indoor agricultural environments. To address these challenges, we introduce HSI-SC-NeRF, a stationary-camera multi-channel NeRF framework for high-throughput hyperspectral 3D reconstruction targeting postharvest inspection of agricultural produce. Multi-view hyperspectral data is captured using a stationary camera while the object rotates within a custom-built Teflon imaging chamber providing diffuse, uniform illumination. Object poses are estimated via ArUco calibration markers and transformed to the camera frame of reference through simulated pose transformations, enabling standard NeRF training on stationary-camera data. A multi-channel NeRF formulation optimizes reconstruction across all hyperspectral bands jointly using a composite spectral loss, supported by a two-stage training protocol that decouples geometric initialization from radiometric refinement. Experiments on three agricultural produce samples demonstrate high spatial reconstruction accuracy and strong spectral fidelity across the visible and near-infrared spectrum, confirming the suitability of HSI-SC-NeRF for integration into automated agricultural workflows. --- ### Beyond Message Passing: A Symbolic Alternative for Expressive and Interpretable Graph Learning Authors: Chuqin Geng, Li Zhang, Haolin Ye, Ziyu Zhao, Yuhe Jiang, Tara Saba, Xinyu Wang, Xujie Si Published: 2026-02-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2602.16947v2 Graph Neural Networks (GNNs) have become essential in high-stakes domains such as drug discovery, yet their black-box nature remains a significant barrier to trustworthiness. While self-explainable GNNs attempt to bridge this gap, they often rely on standard message-passing backbones that inherit fundamental limitations, including the 1-Weisfeiler-Lehman (1-WL) expressivity barrier and a lack of fine-grained interpretability. To address these challenges, we propose SymGraph, a symbolic framework designed to transcend these constraints. By replacing continuous message passing with discrete structural hashing and topological role-based aggregation, our architecture theoretically surpasses the 1-WL barrier, achieving superior expressiveness without the overhead of differentiable optimization. Extensive empirical evaluations demonstrate that SymGraph achieves state-of-the-art performance, outperforming existing self-explainable GNNs. Notably, SymGraph delivers 10x to 100x speedups in training time using only CPU execution. Furthermore, SymGraph generates rules with superior semantic granularity compared to existing rule-based methods, offering great potential for scientific discovery and explainable AI. --- ### Exact Certification of Data-Poisoning Attacks Using Mixed-Integer Programming Authors: Philip Sosnin, Jodie Knapp, Fraser Kennedy, Josh Collyer, Calvin Tsay Published: 2026-02-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2602.16944v1 This work introduces a verification framework that provides both sound and complete guarantees for data poisoning attacks during neural network training. We formulate adversarial data manipulation, model training, and test-time evaluation in a single mixed-integer quadratic programming (MIQCP) problem. Finding the global optimum of the proposed formulation provably yields worst-case poisoning attacks, while simultaneously bounding the effectiveness of all possible attacks on the given training pipeline. Our framework encodes both the gradient-based training dynamics and model evaluation at test time, enabling the first exact certification of training-time robustness. Experimental evaluation on small models confirms that our approach delivers a complete characterization of robustness against data poisoning. --- ### Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents Authors: Arnold Cartagena, Ariane Teixeira Published: 2026-02-18 Categories: cs.AI, cs.SE arXiv: https://arxiv.org/abs/2602.16943v1 Large language models deployed as agents increasingly interact with external systems through tool calls--actions with real-world consequences that text outputs alone do not carry. Safety evaluations, however, overwhelmingly measure text-level refusal behavior, leaving a critical question unanswered: does alignment that suppresses harmful text also suppress harmful actions? We introduce the GAP benchmark, a systematic evaluation framework that measures divergence between text-level safety and tool-call-level safety in LLM agents. We test six frontier models across six regulated domains (pharmaceutical, financial, educational, employment, legal, and infrastructure), seven jailbreak scenarios per domain, three system prompt conditions (neutral, safety-reinforced, and tool-encouraging), and two prompt variants, producing 17,420 analysis-ready datapoints. Our central finding is that text safety does not transfer to tool-call safety. Across all six models, we observe instances where the model's text output refuses a harmful request while its tool calls simultaneously execute the forbidden action--a divergence we formalize as the GAP metric. Even under safety-reinforced system prompts, 219 such cases persist across all six models. System prompt wording exerts substantial influence on tool-call behavior: TC-safe rates span 21 percentage points for the most robust model and 57 for the most prompt-sensitive, with 16 of 18 pairwise ablation comparisons remaining significant after Bonferroni correction. Runtime governance contracts reduce information leakage in all six models but produce no detectable deterrent effect on forbidden tool-call attempts themselves. These results demonstrate that text-only safety evaluations are insufficient for assessing agent behavior and that tool-call safety requires dedicated measurement and mitigation. --- ### SourceBench: Can AI Answers Reference Quality Web Sources? Authors: Hexi Jin, Stephen Liu, Yuheng Li, Simran Malik, Yiying Zhang Published: 2026-02-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2602.16942v1 Large language models (LLMs) increasingly answer queries by citing web sources, but existing evaluations emphasize answer correctness rather than evidence quality. We introduce SourceBench, a benchmark for measuring the quality of cited web sources across 100 real-world queries spanning informational, factual, argumentative, social, and shopping intents. SourceBench uses an eight-metric framework covering content quality (content relevance, factual accuracy, objectivity) and page-level signals (e.g., freshness, authority/accountability, clarity), and includes a human-labeled dataset with a calibrated LLM-based evaluator that matches expert judgments closely. We evaluate eight LLMs, Google Search, and three AI search tools over 3996 cited sources using SourceBench and conduct further experiments to understand the evaluation results. Overall, our work reveals four key new insights that can guide future research in the direction of GenAI and web search. --- ### ConvApparel: A Benchmark Dataset and Validation Framework for User Simulators in Conversational Recommenders Authors: Ofer Meshi, Krisztian Balog, Sally Goldman, Avi Caciularu, Guy Tennenholtz, Jihwan Jeong, Amir Globerson, Craig Boutilier Published: 2026-02-18 Categories: cs.CL arXiv: https://arxiv.org/abs/2602.16938v1 The promise of LLM-based user simulators to improve conversational AI is hindered by a critical "realism gap," leading to systems that are optimized for simulated interactions, but may fail to perform well in the real world. We introduce ConvApparel, a new dataset of human-AI conversations designed to address this gap. Its unique dual-agent data collection protocol -- using both "good" and "bad" recommenders -- enables counterfactual validation by capturing a wide spectrum of user experiences, enriched with first-person annotations of user satisfaction. We propose a comprehensive validation framework that combines statistical alignment, a human-likeness score, and counterfactual validation to test for generalization. Our experiments reveal a significant realism gap across all simulators. However, the framework also shows that data-driven simulators outperform a prompted baseline, particularly in counterfactual validation where they adapt more realistically to unseen behaviors, suggesting they embody more robust, if imperfect, user models. --- ### DeepContext: Stateful Real-Time Detection of Multi-Turn Adversarial Intent Drift in LLMs Authors: Justin Albrethsen, Yash Datta, Kunal Kumar, Sharath Rajasekar Published: 2026-02-18 Categories: cs.AI, cs.ET, cs.LG arXiv: https://arxiv.org/abs/2602.16935v1 While Large Language Model (LLM) capabilities have scaled, safety guardrails remain largely stateless, treating multi-turn dialogues as a series of disconnected events. This lack of temporal awareness facilitates a "Safety Gap" where adversarial tactics, like Crescendo and ActorAttack, slowly bleed malicious intent across turn boundaries to bypass stateless filters. We introduce DeepContext, a stateful monitoring framework designed to map the temporal trajectory of user intent. DeepContext discards the isolated evaluation model in favor of a Recurrent Neural Network (RNN) architecture that ingests a sequence of fine-tuned turn-level embeddings. By propagating a hidden state across the conversation, DeepContext captures the incremental accumulation of risk that stateless models overlook. Our evaluation demonstrates that DeepContext significantly outperforms existing baselines in multi-turn jailbreak detection, achieving a state-of-the-art F1 score of 0.84, which represents a substantial improvement over both hyperscaler cloud-provider guardrails and leading open-weight models such as Llama-Prompt-Guard-2 (0.67) and Granite-Guardian (0.67). Furthermore, DeepContext maintains a sub-20ms inference overhead on a T4 GPU, ensuring viability for real-time applications. These results suggest that modeling the sequential evolution of intent is a more effective and computationally efficient alternative to deploying massive, stateless models. --- ### M-estimation under Two-Phase Multiwave Sampling with Applications to Prediction-Powered Inference Authors: Dan M. Kluger, Stephen Bates Published: 2026-02-18 Categories: stat.ME, math.ST, stat.ML arXiv: https://arxiv.org/abs/2602.16933v2 In two-phase multiwave sampling, inexpensive measurements are collected on a large sample and expensive, more informative measurements are adaptively obtained on subsets of units across multiple waves. Adaptively collecting the expensive measurements can increase efficiency but complicates statistical inference. We give valid estimators and confidence intervals for M-estimation under adaptive two-phase multiwave sampling. We focus on the case where proxies for the expensive variables -- such as predictions from pretrained machine learning models -- are available for all units and propose a Multiwave Predict-Then-Debias estimator that combines proxy information with the expensive, higher-quality measurements to improve efficiency while removing bias. We establish asymptotic linearity and normality and propose asymptotically valid confidence intervals. We also develop an approximately greedy sampling strategy that improves efficiency relative to uniform sampling. Data-based simulation studies support the theoretical results and demonstrate efficiency gains. --- ### RankEvolve: Automating the Discovery of Retrieval Algorithms via LLM-Driven Evolution Authors: Jinming Nian, Fangchen Li, Dae Hoon Park, Yi Fang Published: 2026-02-18 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2602.16932v1 Retrieval algorithms like BM25 and query likelihood with Dirichlet smoothing remain strong and efficient first-stage rankers, yet improvements have mostly relied on parameter tuning and human intuition. We investigate whether a large language model, guided by an evaluator and evolutionary search, can automatically discover improved lexical retrieval algorithms. We introduce RankEvolve, a program evolution setup based on AlphaEvolve, in which candidate ranking algorithms are represented as executable code and iteratively mutated, recombined, and selected based on retrieval performance across 12 IR datasets from BEIR and BRIGHT. RankEvolve starts from two seed programs: BM25 and query likelihood with Dirichlet smoothing. The evolved algorithms are novel, effective, and show promising transfer to the full BEIR and BRIGHT benchmarks as well as TREC DL 19 and 20. Our results suggest that evaluator-guided LLM program evolution is a practical path towards automatic discovery of novel ranking algorithms. --- ### Narrow Fine-Tuning Erodes Safety Alignment in Vision-Language Agents Authors: Idhant Gulati, Shivam Raval Published: 2026-02-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2602.16931v2 Lifelong multimodal agents must continuously adapt to new tasks through post-training, but this creates a fundamental tension between acquiring capabilities and preserving safety alignment. We demonstrate that fine-tuning aligned vision-language models on narrow-domain harmful datasets induces severe emergent misalignment that generalizes broadly across unrelated tasks and modalities. Through experiments on Gemma3-4B, we show that misalignment scales monotonically with LoRA rank, and that multimodal evaluation reveals substantially higher misalignment ($70.71 \pm 1.22$ at $r=128$) than text-only evaluation ($41.19 \pm 2.51$), suggesting that unimodal safety benchmarks may underestimate alignment degradation in vision-language models. Critically, even 10\% harmful data in the training mixture induces substantial alignment degradation. Geometric analysis reveals that harmful behaviors occupy a remarkably low-dimensional subspace, with the majority of misalignment information captured in 10 principal components. To mitigate misalignment, we evaluate two strategies: benign narrow fine-tuning and activation-based steering. While both approaches substantially reduce misalignment, neither completely removes the learned harmful behaviors. Our findings highlight the need for robust continual learning frameworks, as current post-training paradigms may not sufficiently preserve alignment in post-deployment settings. --- ### Say It My Way: Exploring Control in Conversational Visual Question Answering with Blind Users Authors: Farnaz Zamiri Zeraati, Yang Trista Cao, Yuehan Qiao, Hal Daumé, Hernisa Kacorri Published: 2026-02-18 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2602.16930v1 Prompting and steering techniques are well established in general-purpose generative AI, yet assistive visual question answering (VQA) tools for blind users still follow rigid interaction patterns with limited opportunities for customization. User control can be helpful when system responses are misaligned with their goals and contexts, a gap that becomes especially consequential for blind users that may rely on these systems for access. We invite 11 blind users to customize their interactions with a real-world conversational VQA system. Drawing on 418 interactions, reflections, and post-study interviews, we analyze prompting-based techniques participants adopted, including those introduced in the study and those developed independently in real-world settings. VQA interactions were often lengthy: participants averaged 3 turns, sometimes up to 21, with input text typically tenfold shorter than the responses they heard. Built on state-of-the-art LLMs, the system lacked verbosity controls, was limited in estimating distance in space and time, relied on inaccessible image framing, and offered little to no camera guidance. We discuss how customization techniques such as prompt engineering can help participants work around these limitations. Alongside a new publicly available dataset, we offer insights for interaction design at both query and system levels. --- ### Discovering Multiagent Learning Algorithms with Large Language Models Authors: Zun Li, John Schultz, Daniel Hennes, Marc Lanctot Published: 2026-02-18 Categories: cs.GT, cs.AI, cs.MA arXiv: https://arxiv.org/abs/2602.16928v3 Much of the advancement in Multi-Agent Reinforcement Learning (MARL) for imperfect-information games has historically depended on the manual, iterative refinement of algorithmic baselines. Recently, evolutionary coding agents powered by Large Language Models (LLMs) have emerged as powerful tools to automate this discovery process. In this work, we deploy one of such agentic frameworks, AlphaEvolve, to navigate the design spaces of two distinct game-theoretic paradigms: counterfactual regret minimization (CFR) and policy-space response oracles (PSRO). This automated search yielded two algorithms: Volatility-Adaptive Discounted (VAD-) CFR and Smoothed Hybrid Optimistic Regret (SHOR-) PSRO, which are consistently competitive with state-of-the-art human-designed baselines across an 18-game evaluation suite spanning Poker, Goofspiel, Liar's Dice, Blotto, and Battleship variants. However, because the LLM optimizes for fitness on a specific training set, it often constructs highly synergistic, complex mechanisms tailored to those environments. Through systematic ablation studies, we demonstrate that while these mechanisms are tightly coupled, the true driver of generalization lies in a minimal algorithmic core. By distilling the LLM's discoveries down to their most fundamental principles, we produce two minimal solvers: Warm-started Optimistic Predictive (WOP-)CFR and Projection Matching (PM-)PSRO. These distilled versions achieve superior performance on generalization with greatly reduced structural complexity, providing a clear methodology for using LLMs in algorithmic discovery. --- ### Poisson-MNL Bandit: Nearly Optimal Dynamic Joint Assortment and Pricing with Decision-Dependent Customer Arrivals Authors: Junhui Cai, Ran Chen, Qitao Huang, Linda Zhao, Wu Zhu Published: 2026-02-18 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2602.16923v1 We study dynamic joint assortment and pricing where a seller updates decisions at regular accounting/operating intervals to maximize the cumulative per-period revenue over a horizon $T$. In many settings, assortment and prices affect not only what an arriving customer buys but also how many customers arrive within the period, whereas classical multinomial logit (MNL) models assume arrivals as fixed, potentially leading to suboptimal decisions. We propose a Poisson-MNL model that couples a contextual MNL choice model with a Poisson arrival model whose rate depends on the offered assortment and prices. Building on this model, we develop an efficient algorithm PMNL based on the idea of upper confidence bound (UCB). We establish its (near) optimality by proving a non-asymptotic regret bound of order $\sqrt{T\log{T}}$ and a matching lower bound (up to $\log T$). Simulation studies underscore the importance of accounting for the dependency of arrival rates on assortment and pricing: PMNL effectively learns customer choice and arrival models and provides joint assortment-pricing decisions that outperform others that assume fixed arrival rates. --- ### Xray-Visual Models: Scaling Vision models on Industry Scale Data Authors: Shlok Mishra, Tsung-Yu Lin, Linda Wang, Hongli Xu, Yimin Liu, Michael Hsu, Chaitanya Ahuja, Hao Yuan, Jianpeng Cheng, Hong-You Chen, Haoyuan Xu, Chao Li et al. Published: 2026-02-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2602.16918v2 We present Xray-Visual, a unified vision model architecture for large-scale image and video understanding trained on industry-scale social media data. Our model leverages over 15 billion curated image-text pairs and 10 billion video-hashtag pairs from Facebook and Instagram, employing robust data curation pipelines that incorporate balancing and noise suppression strategies to maximize semantic diversity while minimizing label noise. We introduce a three-stage training pipeline that combines self-supervised MAE, semi-supervised hashtag classification, and CLIP-style contrastive learning to jointly optimize image and video modalities. Our architecture builds on a Vision Transformer backbone enhanced with efficient token reorganization (EViT) for improved computational efficiency. Extensive experiments demonstrate that Xray-Visual achieves state-of-the-art performance across diverse benchmarks, including ImageNet for image classification, Kinetics and HMDB51 for video understanding, and MSCOCO for cross-modal retrieval. The model exhibits strong robustness to domain shift and adversarial perturbations. We further demonstrate that integrating large language models as text encoders (LLM2CLIP) significantly enhances retrieval performance and generalization capabilities, particularly in real-world environments. Xray-Visual establishes new benchmarks for scalable, multimodal vision models, while maintaining superior accuracy and computational efficiency. --- ### Coverage-aware Semantic Representation Learning for Underrepresented Visual Concepts Authors: Sakib Ahammed, Xia Cui, Wenqi Lu, Xinqi Fan, Bill Cassidy, Xueli Chen, Moi Hoon Yap Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.16917v2 Modern vision models increasingly rely on rich semantic representations that extend beyond class labels to include descriptive concepts, attributes, and contextual cues. However, semantic concepts are not uniformly represented across classes: a concept may be frequent globally yet remain underrepresented within a class, resulting in low class-concept coverage. We formalize this class-concept phenomenon as Semantic Coverage Imbalance (SCI) and study its relationship with learned semantic representations. We find that lower class-concept coverage is consistently associated with weaker semantic representations. To address this, we introduce Coverage-aware Semantic Representation Learning, which shares more semantic structure for low coverage or uncertain relations while preserving more pair-specific structure for high coverage relations. We instantiate this principle in SemCovNet with Coverage-Calibrated Semantic Sharing (CCSS), a coverage- and uncertainty-aware partial-pooling mechanism. Using a range of vision datasets spanning facial attributes and fine-grained recognition, as well as real-world medical datasets, we show that SCI is widespread, degrades representation quality, and can be mitigated through coverage-aware semantic representation learning. --- ### SemCovNet: Towards Fair and Semantic Coverage-Aware Learning for Underrepresented Visual Concepts Authors: Sakib Ahammed, Xia Cui, Xinqi Fan, Wenqi Lu, Moi Hoon Yap Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.16917v1 Modern vision models increasingly rely on rich semantic representations that extend beyond class labels to include descriptive concepts and contextual attributes. However, existing datasets exhibit Semantic Coverage Imbalance (SCI), a previously overlooked bias arising from the long-tailed semantic representations. Unlike class imbalance, SCI occurs at the semantic level, affecting how models learn and reason about rare yet meaningful semantics. To mitigate SCI, we propose Semantic Coverage-Aware Network (SemCovNet), a novel model that explicitly learns to correct semantic coverage disparities. SemCovNet integrates a Semantic Descriptor Map (SDM) for learning semantic representations, a Descriptor Attention Modulation (DAM) module that dynamically weights visual and concept features, and a Descriptor-Visual Alignment (DVA) loss that aligns visual features with descriptor semantics. We quantify semantic fairness using a Coverage Disparity Index (CDI), which measures the alignment between coverage and error. Extensive experiments across multiple datasets demonstrate that SemCovNet enhances model reliability and substantially reduces CDI, achieving fairer and more equitable performance. This work establishes SCI as a measurable and correctable bias, providing a foundation for advancing semantic fairness and interpretable vision learning. --- ### StereoAdapter-2: Globally Structure-Consistent Underwater Stereo Depth Estimation Authors: Zeyu Ren, Xiang Li, Yiran Wang, Zeyu Zhang, Hao Tang Published: 2026-02-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2602.16915v1 Stereo depth estimation is fundamental to underwater robotic perception, yet suffers from severe domain shifts caused by wavelength-dependent light attenuation, scattering, and refraction. Recent approaches leverage monocular foundation models with GRU-based iterative refinement for underwater adaptation; however, the sequential gating and local convolutional kernels in GRUs necessitate multiple iterations for long-range disparity propagation, limiting performance in large-disparity and textureless underwater regions. In this paper, we propose StereoAdapter-2, which replaces the conventional ConvGRU updater with a novel ConvSS2D operator based on selective state space models. The proposed operator employs a four-directional scanning strategy that naturally aligns with epipolar geometry while capturing vertical structural consistency, enabling efficient long-range spatial propagation within a single update step at linear computational complexity. Furthermore, we construct UW-StereoDepth-80K, a large-scale synthetic underwater stereo dataset featuring diverse baselines, attenuation coefficients, and scattering parameters through a two-stage generative pipeline combining semantic-aware style transfer and geometry-consistent novel view synthesis. Combined with dynamic LoRA adaptation inherited from StereoAdapter, our framework achieves state-of-the-art zero-shot performance on underwater benchmarks with 17% improvement on TartanAir-UW and 7.2% improvment on SQUID, with real-world validation on the BlueROV2 platform demonstrates the robustness of our approach. Code: https://github.com/AIGeeksGroup/StereoAdapter-2. Website: https://aigeeksgroup.github.io/StereoAdapter-2. --- ### A statistical perspective on transformers for small longitudinal cohort data Authors: Kiana Farhadyar, Maren Hackenberg, Kira Ahrens, Charlotte Schenk, Bianca Kollmann, Oliver Tüscher, Klaus Lieb, Michael M. Plichta, Andreas Reif, Raffael Kalisch, Martin Wolkewitz, Moritz Hess et al. Published: 2026-02-18 Categories: stat.ME, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2602.16914v1 Modeling of longitudinal cohort data typically involves complex temporal dependencies between multiple variables. There, the transformer architecture, which has been highly successful in language and vision applications, allows us to account for the fact that the most recently observed time points in an individual's history may not always be the most important for the immediate future. This is achieved by assigning attention weights to observations of an individual based on a transformation of their values. One reason why these ideas have not yet been fully leveraged for longitudinal cohort data is that typically, large datasets are required. Therefore, we present a simplified transformer architecture that retains the core attention mechanism while reducing the number of parameters to be estimated, to be more suitable for small datasets with few time points. Guided by a statistical perspective on transformers, we use an autoregressive model as a starting point and incorporate attention as a kernel-based operation with temporal decay, where aggregation of multiple transformer heads, i.e. different candidate weighting schemes, is expressed as accumulating evidence on different types of underlying characteristics of individuals. This also enables a permutation-based statistical testing procedure for identifying contextual patterns. In a simulation study, the approach is shown to recover contextual dependencies even with a small number of individuals and time points. In an application to data from a resilience study, we identify temporal patterns in the dynamics of stress and mental health. This indicates that properly adapted transformers can not only achieve competitive predictive performance, but also uncover complex context dependencies in small data settings. --- ### A Reversible Semantics for Janus Authors: Ivan Lanese, Germán Vidal Published: 2026-02-18 Categories: cs.PL, cs.AI, cs.LO arXiv: https://arxiv.org/abs/2602.16913v2 Janus is a paradigmatic example of a reversible programming language. Indeed, Janus programs can be executed backwards as well as forwards. However, its current small-step semantics (useful, e.g., for debugging or as a basis for extensions with concurrency primitives) is not reversible, since it loses information while computing forwards. E.g., it does not satisfy the Loop Lemma, stating that any reduction has an inverse, a main property of reversibility in process calculi, where a small-step semantics is commonly used. We present here a novel small-step semantics which is actually reversible, while remaining equivalent to the previous one. It involves the non-trivial challenge of defining a semantics based on a "program counter" for a high-level programming language. --- ### Multi-objective optimization and quantum hybridization of equivariant deep learning interatomic potentials Authors: G. Laskaris, D. Morozov, D. Tarpanov, A. Seth, J. Procelewska, G. Sai Gautam, A. Sagingalieva, R. Brasher, A. Melnikov Published: 2026-02-18 Categories: cond-mat.mtrl-sci, cs.LG, quant-ph arXiv: https://arxiv.org/abs/2602.16908v2 Allegro is a machine learning interatomic potential model designed to predict atomic properties in molecules using E(3) equivariant neural networks. When training this model, there tends to be a trade-off between accuracy and inference time. For this reason, we apply multi-objective hyperparameter optimization to both objectives. Additionally, we experiment with modified architectures by constructing variants of Allegro: one extended with additional classical layers and one incorporating quantum-classical hybrid layers. We evaluate all models on QM9, rMD17-aspirin, rMD17-benzene, and a self-generated dataset of copper-lithium structures. As results, both variants surpass Allegro in force prediction accuracy across multiple datasets. The classical variant consistently improves over the baseline, while the quantum-classical hybrid variant achieves the best overall force prediction accuracy on the Cu-Li dataset, where it was fully optimized, outperforming the classical variant by approximately 13%. Notably, the hybrid variant also achieves competitive results on the remaining datasets despite using hyperparameters transferred from Cu-Li without dataset-specific optimization, suggesting that quantum-classical hybridization is a promising direction for enhancing MLIP architectures. --- ### LLM-WikiRace Benchmark: How Far Can LLMs Plan over Real-World Knowledge Graphs? Authors: Juliusz Ziomek, William Bankes, Lorenz Wolf, Shyam Sundhar Ramesh, Xiaohang Tang, Ilija Bogunovic Published: 2026-02-18 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2602.16902v5 We introduce LLM-Wikirace, a benchmark for evaluating planning, reasoning, and world knowledge in large language models (LLMs). In LLM-Wikirace, models must efficiently navigate Wikipedia hyperlinks step by step to reach a target page from a given source, requiring look-ahead planning and the ability to reason about how concepts are connected in the real world. We evaluate a broad set of open- and closed-source models, including Gemini-3, GPT-5, and Claude Opus 4.5, which achieve the strongest results on the easy level of the task and demonstrate superhuman performance. Despite this, performance drops sharply on hard difficulty: the best-performing model, Gemini-3, succeeds in only 23\% of hard games, highlighting substantial remaining challenges for frontier models. Our analysis shows that world knowledge is a necessary ingredient for success, but only up to a point, beyond this threshold, planning and long-horizon reasoning capabilities become the dominant factors. Trajectory-level analysis further reveals that even the strongest models struggle to replan after failure, frequently entering loops rather than recovering. LLM-Wikirace is a simple benchmark that reveals clear limitations in current reasoning systems, offering an open arena where planning-capable LLMs still have much to prove. Our code and leaderboard available at https:/llmwikirace.github.io. --- ### AgentLAB: Benchmarking LLM Agents against Long-Horizon Attacks Authors: Tanqiu Jiang, Yuhui Wang, Jiacheng Liang, Ting Wang Published: 2026-02-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2602.16901v1 LLM agents are increasingly deployed in long-horizon, complex environments to solve challenging problems, but this expansion exposes them to long-horizon attacks that exploit multi-turn user-agent-environment interactions to achieve objectives infeasible in single-turn settings. To measure agent vulnerabilities to such risks, we present AgentLAB, the first benchmark dedicated to evaluating LLM agent susceptibility to adaptive, long-horizon attacks. Currently, AgentLAB supports five novel attack types including intent hijacking, tool chaining, task injection, objective drifting, and memory poisoning, spanning 28 realistic agentic environments, and 644 security test cases. Leveraging AgentLAB, we evaluate representative LLM agents and find that they remain highly susceptible to long-horizon attacks; moreover, defenses designed for single-turn interactions fail to reliably mitigate long-horizon threats. We anticipate that AgentLAB will serve as a valuable benchmark for tracking progress on securing LLM agents in practical settings. The benchmark is publicly available at https://tanqiujiang.github.io/AgentLAB_main. --- ### MALLVI: A Multi-Agent Framework for Integrated Generalized Robotics Manipulation Authors: Mehrshad Taji, Arad Mahdinezhad Kashani, Iman Ahmadi, AmirHossein Jadidi, Saina Kashani, Babak Khalaj Published: 2026-02-18 Categories: cs.RO, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2602.16898v6 Task planning for robotic manipulation with large language models (LLMs) is an emerging area. Prior approaches rely on specialized models, fine tuning, or prompt tuning, and often operate in an open loop manner without robust environmental feedback, making them fragile in dynamic settings. MALLVI presents a Multi Agent Large Language and Vision framework that enables closed-loop feedback driven robotic manipulation. Given a natural language instruction and an image of the environment, MALLVI generates executable atomic actions for a robot manipulator. After action execution, a Vision Language Model (VLM) evaluates environmental feedback and decides whether to repeat the process or proceed to the next step. Rather than using a single model, MALLVI coordinates specialized agents, Decomposer, Localizer, Thinker, and Reflector, to manage perception, localization, reasoning, and high level planning. An optional Descriptor agent provides visual memory of the initial state. The Reflector supports targeted error detection and recovery by reactivating only relevant agents, avoiding full replanning. Experiments in simulation and real-world settings show that iterative closed loop multi agent coordination improves generalization and increases success rates in zero shot manipulation tasks. Code available at https://github.com/iman1234ahmadi/MALLVI . ---