Humans subconsciously choose robust ways of selecting and using tools, for example, choosing a ladle over a flat spatula to serve meatballs. However, robustness under external disturbances remains underexplored in robotic tool-use planning. This paper presents a robustness-aware method that jointly selects tools and plans contact-rich manipulation trajectories, explicitly optimizing for robustness against disturbances. At the core of our method is an energy-based robustness metric that guides the planner toward robust manipulation behaviors. We formulate a hierarchical optimization pipeline that first identifies a tool and configuration that optimizes robustness, and then plans a corresponding manipulation trajectory that maintains robustness throughout execution. We evaluate our method across three representative tool-use tasks. Simulation and real-world results demonstrate that our method consistently selects robust tools and generates disturbance-resilient manipulation plans.

We introduce ESGenius, a comprehensive benchmark for evaluating and enhancing the proficiency of Large Language Models (LLMs) in Environmental, Social, and Governance (ESG) and sustainability-focused question answering. ESGenius comprises two key components: (i) ESGenius-QA, a collection of 1,136 Multiple-Choice Questions (MCQs) generated by LLMs and rigorously validated by domain experts, covering a broad range of ESG pillars and sustainability topics. Each question is systematically linked to its corresponding source text, enabling transparent evaluation and supporting Retrieval-Augmented Generation (RAG) methods; and (ii) ESGenius-Corpus, a meticulously curated repository of 231 foundational frameworks, standards, reports, and recommendation documents from 7 authoritative sources. Moreover, to fully assess the capabilities and adaptation potential of LLMs, we implement a rigorous two-stage evaluation protocol -- Zero-Shot and RAG. Extensive experiments across 50 LLMs (0.5B to 671B) demonstrate that state-of-the-art models achieve only moderate performance in zero-shot settings, with accuracies around 55--70%, highlighting a significant knowledge gap for LLMs in this specialized, interdisciplinary domain. However, models employing RAG demonstrate significant performance improvements, particularly for smaller models. For example, DeepSeek-R1-Distill-Qwen-14B improves from 63.82% (zero-shot) to 80.46% with RAG. These results demonstrate the necessity of grounding responses in authoritative sources for enhanced ESG understanding. To the best of our knowledge, ESGenius is the first comprehensive QA benchmark designed to rigorously evaluate LLMs on ESG and sustainability knowledge, providing a critical tool to advance trustworthy AI in this vital domain.

Deep learning based Image Super-Resolution (ISR) relies on large training datasets to optimize model generalization; this requires substantial computational and storage resources during training. While dataset condensation (DC) has shown potential in improving data efficiency for high-level computer vision tasks, adopting these methods for ISR is not straightforward due to the different requirements of ISR training, including the use of unlabeled datasets and high resolution images with fine details. In this paper, we propose a novel Instance Data Condensation (IDC) framework specifically for ISR, which achieves data condensation in a per-image manner, aiming to address the limitations when directly applying existing DC methods to the ISR task. Furthermore, the IDC framework is based on a novel Random Local Fourier Feature Extraction and Multi-level Feature Distribution Matching methods, which are designed to generate high-quality synthesized content by aligning its feature distributions with those of the original high-resolution training samples at both global and local levels. This framework has been utilized to condense the most commonly used training dataset for ISR, DIV2K, with a 10% condensation rate. The resulting synthetic dataset offers comparable performance to the original full dataset and excellent training stability when used to train various popular ISR models. To the best of our knowledge, this is the first time that a condensed/synthetic dataset (with a 10% data volume) has demonstrated such performance.

Pre-training equips text-to-image (T2I) models with broad world knowledge, but this alone is often insufficient to achieve high aesthetic quality and alignment. Consequently, supervised fine-tuning (SFT) is crucial for further refinement. However, its effectiveness highly depends on the quality of the fine-tuning dataset. Existing public SFT datasets frequently target narrow domains (e.g., anime or specific art styles), and the creation of high-quality, general-purpose SFT datasets remains a significant challenge. Current curation methods are often costly and struggle to identify truly impactful samples. This challenge is further complicated by the scarcity of public general-purpose datasets, as leading models often rely on large, proprietary, and poorly documented internal data, hindering broader research progress. This paper introduces a novel methodology for creating general-purpose SFT datasets by leveraging a pre-trained generative model as an estimator of high-impact training samples. We apply this methodology to construct and release Alchemist, a compact (3,350 samples) yet highly effective SFT dataset. Experiments demonstrate that Alchemist substantially improves the generative quality of five public T2I models while preserving diversity and style. Additionally, we release the fine-tuned models' weights to the public.

Diffusion Transformers (DiTs) have emerged as the state-of-the-art architecture for video generation, yet their computational and memory demands hinder practical deployment. While post-training quantization (PTQ) presents a promising approach to accelerate Video DiT models, existing methods suffer from two critical limitations: (1) dependence on computation-heavy and inflexible calibration procedures, and (2) considerable performance deterioration after quantization. To address these challenges, we propose DVD-Quant, a novel Data-free quantization framework for Video DiTs. Our approach integrates three key innovations: (1) Bounded-init Grid Refinement (BGR) and (2) Auto-scaling Rotated Quantization (ARQ) for calibration data-free quantization error reduction, as well as (3) $δ$-Guided Bit Switching ($δ$-GBS) for adaptive bit-width allocation. Extensive experiments across multiple video generation benchmarks demonstrate that DVD-Quant achieves an approximately 2$\times$ speedup over full-precision baselines on advanced DiT models while maintaining visual fidelity. Notably, DVD-Quant is the first to enable W4A4 PTQ for Video DiTs without compromising video quality. Code and models will be available at https://github.com/lhxcs/DVD-Quant.

The paper presents a novel sample-based algorithm, called C*, for real-time coverage path planning (CPP) of unknown environments. C* is built upon the concept of a Rapidly Covering Graph (RCG), which is incrementally constructed during robot navigation via progressive sampling of the search space. By using efficient sampling and pruning techniques, the RCG is constructed to be a minimum-sufficient graph, where its nodes and edges form the potential waypoints and segments of the coverage trajectory, respectively. The RCG tracks the coverage progress, generates the coverage trajectory and helps the robot to escape from the dead-end situations. To minimize coverage time, C* produces the desired back-and-forth coverage pattern, while adapting to the TSP-based optimal coverage of local isolated regions, called coverage holes, which are surrounded by obstacles and covered regions. It is analytically proven that C* provides complete coverage of unknown environments. The algorithmic simplicity and low computational complexity of C* make it easy to implement and suitable for real-time on-board applications. The performance of C* is validated by 1) extensive high-fidelity simulations and 2) laboratory experiments using an autonomous robot. C* yields near optimal trajectories, and a comparative evaluation with seven existing CPP methods demonstrates significant improvements in performance in terms of coverage time, number of turns, trajectory length, and overlap ratio, while preventing the formation of coverage holes. Finally, C* is comparatively evaluated on two different CPP applications using 1) energy-constrained robots and 2) multi-robot teams.

Event cameras deliver visual data with high temporal resolution, low latency, and minimal redundancy, yet their asynchronous, sparse sequential nature challenges standard tensor-based machine learning (ML). While the recent asynchronous-to-synchronous (A2S) paradigm aims to bridge this gap by asynchronously encoding events into learned features for ML pipelines, existing A2S approaches often sacrifice expressivity and generalizability compared to dense, synchronous methods. This paper introduces EVA (EVent Asynchronous feature learning), a novel A2S framework to generate highly expressive and generalizable event-by-event features. Inspired by the analogy between events and language, EVA uniquely adapts advances from language modeling in linear attention and self-supervised learning for its construction. In demonstration, EVA outperforms prior A2S methods on recognition tasks (DVS128-Gesture and N-Cars), and represents the first A2S framework to successfully master demanding detection tasks, achieving a 0.477 mAP on the Gen1 dataset. These results underscore EVA's potential for advancing real-time event-based vision applications.

Reward modeling is essential for aligning large language models with human preferences through reinforcement learning. To provide accurate reward signals, a reward model (RM) should stimulate deep thinking and conduct interpretable reasoning before assigning a score or a judgment. Inspired by recent advances of long chain-of-thought on reasoning-intensive tasks, we hypothesize and validate that integrating reasoning into reward modeling significantly enhances RM's interpretability and performance. We introduce a new class of generative reward models, Reasoning Reward Models (ReasRMs), which formulate reward modeling as a reasoning task. We propose a reasoning-oriented training pipeline and train a family of ReasRMs, RM-R1. RM-R1 features a chain-of-rubrics (CoR) mechanism -- self-generating sample-level chat rubrics or math/code solutions, and evaluating candidate responses against them. The training of RM-R1 consists of two key stages: (1) distillation of high-quality reasoning chains and (2) reinforcement learning with verifiable rewards. Empirically, our models achieve superior performance across three reward model benchmarks on average, outperforming much larger open-weight models (e.g., INF-ORM-Llama3.1-70B) and proprietary ones (e.g., GPT-4o) by up to 4.9%. Beyond final performance, we perform thorough analyses to understand the key ingredients of successful ReasRM training.

The Boltzmann equation, a fundamental model in kinetic theory, describes the evolution of particle distribution functions through a nonlinear, high-dimensional collision operator. However, its numerical solution remains computationally demanding, particularly for inelastic collisions and high-dimensional velocity domains. In this work, we propose the Fourier Neural Spectral Network (FourierSpecNet), a hybrid framework that integrates the Fourier spectral method with deep learning to approximate the collision operator in Fourier space efficiently. FourierSpecNet achieves resolution-invariant learning and supports zero-shot super-resolution, enabling accurate predictions at unseen resolutions without retraining. Beyond empirical validation, we establish a consistency result showing that the trained operator converges to the spectral solution as the discretization is refined. We evaluate our method on several benchmark cases, including Maxwellian and hard-sphere molecular models, as well as inelastic collision scenarios. The results demonstrate that FourierSpecNet offers competitive accuracy while significantly reducing computational cost compared to traditional spectral solvers. Our approach provides a robust and scalable alternative for solving the Boltzmann equation across both elastic and inelastic regimes.

Federated Learning (FL) has emerged as a transformative paradigm in the field of distributed machine learning, enabling multiple clients such as mobile devices, edge nodes, or organizations to collaboratively train a shared global model without the need to centralize sensitive data. This decentralized approach addresses growing concerns around data privacy, security, and regulatory compliance, making it particularly attractive in domains such as healthcare, finance, and smart IoT systems. This survey provides a concise yet comprehensive overview of Federated Learning, beginning with its core architecture and communication protocol. We discuss the standard FL lifecycle, including local training, model aggregation, and global updates. A particular emphasis is placed on key technical challenges such as handling non-IID (non-independent and identically distributed) data, mitigating system and hardware heterogeneity, reducing communication overhead, and ensuring privacy through mechanisms like differential privacy and secure aggregation. Furthermore, we examine emerging trends in FL research, including personalized FL, cross-device versus cross-silo settings, and integration with other paradigms such as reinforcement learning and quantum computing. We also highlight real-world applications and summarize benchmark datasets and evaluation metrics commonly used in FL research. Finally, we outline open research problems and future directions to guide the development of scalable, efficient, and trustworthy FL systems.

Zero-shot anomaly detection (ZSAD) aims to identify anomalies in unseen categories by leveraging CLIP's zero-shot capabilities to match text prompts with visual features. A key challenge in ZSAD is learning general prompts stably and utilizing them effectively, while maintaining both generalizability and category specificity. Although general prompts have been explored in prior works, achieving their stable optimization and effective deployment remains a significant challenge. In this work, we propose GenCLIP, a novel framework that learns and leverages general prompts more effectively through multi-layer prompting and dual-branch inference. Multi-layer prompting integrates category-specific visual cues from different CLIP layers, enriching general prompts with more comprehensive and robust feature representations. By combining general prompts with multi-layer visual features, our method further enhances its generalization capability. To balance specificity and generalization, we introduce a dual-branch inference strategy, where a vision-enhanced branch captures fine-grained category-specific features, while a query-only branch prioritizes generalization. The complementary outputs from both branches improve the stability and reliability of anomaly detection across unseen categories. Additionally, we propose an adaptive text prompt filtering mechanism, which removes irrelevant or atypical class names not encountered during CLIP's training, ensuring that only meaningful textual inputs contribute to the final vision-language alignment.

As Large Language Models (LLMs) are deployed across diverse regions, aligning them with pluralistic cultures is crucial for improving user engagement and mitigating cultural conflicts. Recent work has curated, either synthesized or manually annotated, culture-specific corpora for alignment. Nevertheless, inspired by cultural theories, we recognize they face two key challenges. (1) Representativeness: These corpora inadequately capture the target culture's core characteristics, causing insufficient cultural coverage and redundancy; (2) Distinctiveness: They fail to distinguish the unique nuances of the target culture from patterns shared across relevant ones, hindering precise culture modeling. To handle these challenges, we introduce CAReDiO, a novel data optimization framework that alternately optimizes culture-sensitive questions and responses according to two information-theoretic objectives in an in-context manner, enhancing both cultural representativeness and distinctiveness of constructed data. Extensive experiments on 15 cultures demonstrate that CAReDiO can create high-quality data with richer cultural information and enable efficient alignment of small open-source or large proprietary LLMs with as few as 200 training samples, consistently outperforming previous datasets in both multi-choice and open-ended benchmarks.

Using pre-trained large language models (LLMs) as a backbone for time series prediction has recently attracted growing research interest. Existing approaches typically split time series into patches, map them to the token space of LLMs via a Tokenizer, process the tokens through a frozen or fine-tuned LLM backbone, and then reconstruct numerical forecasts using a Detokenizer. However, the actual effectiveness of LLMs for time series forecasting remains under debate. We observe that when trained and evaluated on small datasets, these Tokenizer-Detokenizer pairs often overfit to the specific data distribution, thereby masking the intrinsic predictive capability of the LLM backbone. To investigate the inherent potential of LLMs in this context, we design three models with identical architectures but distinct pre-training strategies. By leveraging large-scale pre-training, we obtain more unbiased Tokenizer-Detokenizer pairs that are seamlessly integrated with the LLM backbone. Through controlled experiments, we evaluate the zero-shot and few-shot forecasting performance of the LLM, offering insights into its true capabilities. Our extensive experiments reveal that, although the LLM backbone shows some promise, its performance remains limited and does not consistently surpass that of models specifically trained on large-scale time series data. Our source code is publicly available in the repository: https://github.com/SiriZhang45/LLM4TS.

Geometrically accurate and semantically expressive map representations have proven invaluable for robot deployment and task planning in unknown environments. Nevertheless, real-time, open-vocabulary semantic understanding of large-scale unknown environments still presents open challenges, mainly due to computational requirements. In this paper we present FindAnything, an open-world mapping framework that incorporates vision-language information into dense volumetric submaps. Thanks to the use of vision-language features, FindAnything combines pure geometric and open-vocabulary semantic information for a higher level of understanding. It proposes an efficient storage of open-vocabulary information through the aggregation of features at the object level. Pixelwise vision-language features are aggregated based on eSAM segments, which are in turn integrated into object-centric volumetric submaps, providing a mapping from open-vocabulary queries to 3D geometry that is scalable also in terms of memory usage. We demonstrate that FindAnything performs on par with the state-of-the-art in terms of semantic accuracy while being substantially faster and more memory-efficient, allowing its deployment in large-scale environments and on resourceconstrained devices, such as MAVs. We show that the real-time capabilities of FindAnything make it useful for downstream tasks, such as autonomous MAV exploration in a simulated Search and Rescue scenario. Project Page: https://ethz-mrl.github.io/findanything/.

Multi-modal music generation, using multiple modalities like text, images, and video alongside musical scores and audio as guidance, is an emerging research area with broad applications. This paper reviews this field, categorizing music generation systems from the perspective of modalities. The review covers modality representation, multi-modal data alignment, and their utilization to guide music generation. Current datasets and evaluation methods are also discussed. Key challenges in this area include effective multi-modal integration, large-scale comprehensive datasets, and systematic evaluation methods. Finally, an outlook on future research directions is provided, focusing on creativity, efficiency, multi-modal alignment, and evaluation.

Lidar-only odometry aims to estimate the trajectory of a mobile platform from a stream of lidar scans. Traditional scan-to map approaches register each scan against a single, evolving map, which propagates registration errors over time. To mitigate this, we propose a multitude-of-maps approach where the current scan is registered against multiple overlapping submaps instead of a single static map. By optimizing the resulting constraints in a pose graph, our method enables not only precise estimation of the current pose but also retrospective refinement of the submaps' anchor points, which improves short-term consistency and long-term accuracy. We demonstrate that our approach achieves competitive and often superior accuracy on a variety of automotive datasets while maintaining real-time performance. Ablation studies confirm the critical role of multiple registrations and retrospective refinement of the map as core factors for our accuracy gains. Code and raw results are available on our public GitHub at https://github.com/Fusion-Goettingen/IROS_2026_Kurda_Graph.

Surficial geologic (SG) maps are essential for understanding surface processes and supporting infrastructure planning, but current workflows are labor-intensive and difficult to scale. We introduce EarthScape, an AI-ready multimodal dataset for SG mapping that integrates digital elevation models, aerial imagery, multi-scale terrain features, and hydrologic and infrastructure vector data within a unified, reproducible pipeline. We report baseline benchmarks across single-modality, multi-scale, and multimodal configurations. Our experiments show that terrain features provide the most reliable predictive signal, while raw spectral and elevation inputs degrade substantially under cross-region evaluation. EarthScape offers a geographically compact, but modality-rich benchmark for multimodal fusion, domain adaptation, and surface modeling. EarthScape is available for direct download at https://uknowledge.uky.edu/kgs_data/16/, and code is available at https://github.com/masseygeo/earthscape.

Flow-based Transformer models have achieved state-of-the-art image generation performance, but often suffer from high inference latency and computational cost due to their large parameter sizes. To improve inference efficiency without compromising quality, we propose Bridged Progressive Rectified Flow Transformers (NAMI), which decompose the generation process across temporal, spatial, and architectural demensions. We divide the rectified flow into different stages according to resolution, and use a BridgeFlow module to connect them. Fewer Transformer layers are used at low-resolution stages to generate image layouts and concept contours, and more layers are progressively added as the resolution increases. Experiments demonstrate that our approach achieves fast convergence and reduces inference time while ensuring generation quality. The main contributions of this paper are summarized as follows: (1) We introduce Bridged Progressive Rectified Flow Transformers that enable multi-resolution training, accelerating model convergence; (2) NAMI leverages piecewise flow and spatial cascading of Diffusion Transformer (DiT) to rapidly generate images, reducing inference time by 64% for generating 1024 resolution images; (3) We propose a BridgeFlow module to align flows between different stages; (4) We propose the NAMI-1K benchmark to evaluate human preference performance, aiming to mitigate distributional bias and comprehensively assess model effectiveness. The results show that our model is competitive with state-of-the-art models.

The reliance on large-scale datasets and extensive computational resources has become a major barrier to advancing representation learning in vision, especially in data-scarce domains. In this paper, we address the critical question: Can we escape the big data paradigm in self-supervised representation learning from images? We introduce SCOTT (Sparse Convolutional Tokenizer for Transformers), a shallow tokenization architecture that is compatible with Masked Image Modeling (MIM) tasks. SCOTT injects convolutional inductive biases into Vision Transformers (ViTs), enhancing their efficacy in small-scale data regimes. Alongside, we propose to train on a Joint-Embedding Predictive Architecture within a MIM framework (MIM-JEPA), operating in latent representation space to capture more semantic features. Our approach enables ViTs to be trained from scratch on datasets orders of magnitude smaller than traditionally required --without relying on massive external datasets for pretraining. We validate our method on three small-size, standard-resoultion, fine-grained datasets: Oxford Flowers-102, Oxford IIIT Pets-37, and ImageNet-100. Despite the challenges of limited data and high intra-class similarity, frozen SCOTT models pretrained with MIM-JEPA significantly outperform fully supervised methods and achieve competitive results with SOTA approaches that rely on large-scale pretraining, complex image augmentations and bigger model sizes. By demonstrating that robust off-the-shelf representations can be learned with limited data, compute, and model sizes, our work paves the way for computer applications in resource constrained environments such as medical imaging or robotics. Our findings challenge the prevailing notion that vast amounts of data are indispensable for effective representation learning in vision, offering a new pathway toward more accessible and inclusive advancements in the field.

Compound AI systems, comprising multiple interacting components such as LLMs, foundation models, and external tools, have demonstrated remarkable improvements compared to single models in various tasks. To ensure their effective deployment in real-world applications, aligning these systems with human preferences is crucial. However, aligning the compound system via policy optimization, unlike the alignment of a single model, is challenging for two main reasons: (i) non-differentiable interactions between components make end-to-end gradient-based optimization method inapplicable, and (ii) system-level preferences cannot be directly transformed into component-level preferences. To address these challenges, we first formulate compound AI systems as Directed Acyclic Graphs (DAGs), explicitly modeling both component interactions and the associated data flows. Building on this formulation, we introduce $\textbf{SysDPO}$, a framework that extends Direct Preference Optimization (DPO) to enable joint system-level alignment. We propose two variants, SysDPO-Direct and SysDPO-Sampling, tailored for scenarios depending on whether we construct a system-specific preference dataset. We empirically demonstrate the effectiveness of our approach across two applications: the joint alignment of a language model and a diffusion model, and the joint alignment of an LLM collaboration system.

We introduce FragFM, a novel hierarchical framework via fragment-level discrete flow matching for efficient molecular graph generation. FragFM generates molecules at the fragment level, leveraging a coarse-to-fine autoencoder to reconstruct details at the atom level. Together with a stochastic fragment bag strategy to effectively handle a large fragment space, our framework enables more efficient, scalable molecular generation. We demonstrate that our fragment-based approach achieves better property control than the atom-based method and additional flexibility through conditioning the fragment bag. We also propose a Natural Product Generation benchmark (NPGen) to evaluate the ability of modern molecular graph generative models to generate natural product-like molecules. Since natural products are biologically prevalidated and differ from typical drug-like molecules, our benchmark provides a more challenging yet meaningful evaluation relevant to drug discovery. We conduct a comparative study of FragFM against various models on diverse molecular generation benchmarks, including NPGen, demonstrating superior performance. The results highlight the potential of fragment-based generative modeling for large-scale, property-aware molecular design, paving the way for more efficient exploration of chemical space.

Generative control policies have recently unlocked major progress in robotics. These methods produce action sequences via diffusion or flow matching, with training data provided by demonstrations. But existing methods come with two key limitations: they require expert demonstrations, which can be difficult to obtain, and they are limited to relatively slow, quasi-static tasks. In this paper, we leverage a tight connection between sampling-based predictive control and generative modeling to address each of these issues. In particular, we introduce generative predictive control, a supervised learning framework for tasks with fast dynamics that are easy to simulate but difficult to demonstrate. We then show how trained flow-matching policies can be warm-started at inference time, maintaining temporal consistency and enabling high-frequency feedback. We believe that generative predictive control offers a complementary approach to existing behavior cloning methods, and hope that it paves the way toward generalist policies that extend beyond quasi-static demonstration-oriented tasks.

Code-switching (CS) is still a critical challenge in Natural Language Processing (NLP), due to the limited availability of large-scale, diverse CS datasets for robust training and evaluation. Despite recent advances, the capabilities and limitations of LLMs in handling CS are still not fully understood. In this work, we investigate the extent to which LLMs can be used in a framework for CS text generation, focusing on the English-Spanish language pair. Our proposed methodology consists of back-translating natural CS sentences into monolingual English, and using the resulting parallel corpus to fine-tune LLMs to turn monolingual sentences into CS. We thoroughly analyse the models' performance through a study on human preferences, a qualitative error analysis, an evaluation with popular reference-based metrics and LLM-based judgment. Results show that fine-tuning can be a key step to ensure that current LLMs consistently generate fluent code-switched text and that our methodology generates high-quality outputs, expanding research opportunities in CS communication. We find that traditional metrics do not correlate with human judgement when assessing the quality of the generated CS data, but LLM-based judgment aligns more closely with human preferences. We release our code and generated dataset under a CC-BY-NC-SA license.

With the advent of large multimodal language models, science is now at a threshold of an AI-based technological transformation. An emerging ecosystem of models and tools aims to support researchers throughout the scientific lifecycle, including (1) searching for relevant literature, (2) generating research ideas and conducting experiments, (3) producing text-based content, (4) creating multimodal artifacts such as figures and diagrams, and (5) evaluating scientific work, as in peer review. In this survey, we provide a curated overview of literature representative of the core techniques, evaluation practices, and emerging trends in AI-assisted scientific discovery. Across the five tasks outlined above, we discuss datasets, methods, results, evaluation strategies, limitations, and ethical concerns, including risks to research integrity through the misuse of generative models. We aim for this survey to serve both as an accessible, structured orientation for newcomers to the field, as well as a catalyst for new AI-based initiatives and their integration into future ``AI4Science'' systems.

Neural View Synthesis (NVS) techniques such as NeRF and 3D Gaussian Splatting (3DGS) have enabled photorealistic rendering from novel viewpoints and are increasingly used to augment training data for visual localization. However, these methods fundamentally rely on observed geometry and radiance; they interpolate existing information but cannot hallucinate unseen 3D structures or recover missing content under sparse or extreme viewpoints. As a result, rendered views often exhibit blur, structural distortion, or incomplete geometry. While such imperfections may be tolerated by Camera Pose Regression (CPR) methods, they severely degrade Scene Coordinate Regression (SCR), which requires accurate per-pixel 3D supervision. To address this limitation, we introduce PoI (Pixel-of-Interest), a framework that enables effective NVS augmentation for SCR-based localization. We first employ 3DGS to render novel views and leverage a single-step diffusion model to refine them, allowing the synthesis of structurally plausible details beyond purely geometry-driven interpolation. However, even diffusion-refined views may contain unreliable pixels. Therefore, we propose a progressive pixel-level filtering strategy based on reprojection error to selectively retain trustworthy synthetic pixels during training while suppressing harmful ones. Extensive experiments on 7Scenes and Cambridge Landmarks demonstrate that our method consistently improves localization accuracy over strong SCR baselines and achieves state-of-the-art performance with competitive training efficiency. Our results reveal that, for SCR, the benefit of novel view augmentation depends not only on generative realism but also on explicit control of pixel-level reliability.

3D Gaussian Splatting (3DGS) has emerged as a powerful approach for 3D scene reconstruction using 3D Gaussians. However, neither the centers nor surfaces of the Gaussians are accurately aligned to the object surface, complicating their direct use in point cloud and mesh reconstruction. Additionally, 3DGS typically produces floater artifacts, increasing the number of Gaussians and storage requirements. To address these issues, we present FeatureGS, which incorporates an additional geometric loss term based on an eigenvalue-derived 3D shape feature into the optimization process of 3DGS. The goal is to improve geometric accuracy and enhance properties of planar surfaces with reduced structural entropy in local 3D neighborhoods.We present four alternative formulations for the geometric loss term based on 'planarity' of Gaussians, as well as 'planarity', 'omnivariance', and 'eigenentropy' of Gaussian neighborhoods. We provide quantitative and qualitative evaluations on 15 scenes of the DTU benchmark dataset focusing on following key aspects: Geometric accuracy and artifact-reduction, measured by the Chamfer distance, and memory efficiency, evaluated by the total number of Gaussians. Additionally, rendering quality is monitored by Peak Signal-to-Noise Ratio. FeatureGS achieves a 30 % improvement in geometric accuracy, reduces the number of Gaussians by 90 %, and suppresses floater artifacts, while maintaining comparable photometric rendering quality. The geometric loss with 'planarity' from Gaussians provides the highest geometric accuracy, while 'omnivariance' in Gaussian neighborhoods reduces floater artifacts and number of Gaussians the most. This makes FeatureGS a strong method for geometrically accurate, artifact-reduced and memory-efficient 3D scene reconstruction, enabling the direct use of Gaussian centers for geometric representation.

The syntactic structure of a sentence can be described as a tree that indicates the syntactic relationships between words. In spite of significant progress in unsupervised methods that retrieve the syntactic structure of sentences, guessing the right direction of edges is still a challenge. As in a syntactic dependency structure edges are oriented away from the root, the challenge of guessing the right direction can be reduced to finding an undirected tree and the root. The limited performance of current unsupervised methods demonstrates the lack of a proper understanding of what a root vertex is from first principles. We consider an ensemble of centrality scores, some that only take into account the free tree (non-spatial scores) and others that take into account the position of vertices (spatial scores). We test the hypothesis that the root vertex is an important or central vertex of the syntactic dependency structure. We confirm the hypothesis in the sense that root vertices tend to have high centrality and that vertices of high centrality tend to be roots. The best performance in guessing the root is achieved by novel scores that only take into account the position of a vertex and that of its neighbours. We provide theoretical and empirical foundations towards a universal notion of rootness from a network science perspective.

Kernel-free quadratic surface support vector machines (QSVM) have recently gained traction due to their flexibility in modeling nonlinear decision boundaries without relying on kernel functions. However, the introduction of a full quadratic classifier significantly increases the number of model parameters, scaling quadratically with data dimensionality, which often leads to overfitting and makes interpretation difficult. To address these challenges, we propose sparse variants of the QSVM by enforcing a cardinality constraint on the model parameters. While enhancing generalization and promoting sparsity, leveraging the $\ell_0$-norm inevitably incurs additional computational complexity. To tackle this, we develop a penalty decomposition algorithm capable of producing solutions that provably satisfy the first-order Lu-Zhang optimality conditions. We show that the subproblems arising within the algorithm either admit closed-form solutions or can be solved efficiently through dual formulations, which contributes to the method's overall effectiveness. Besides, we analyze the convergence behavior of the algorithm under both loss settings. In addition, the numerical experiments on public benchmark datasets indicate that the proposed model is competitive with commonly used SVM variants and produces sparse solutions as expected. Moreover, its strong performance on real-world credit datasets demonstrates its potential for credit scoring applications.

Mixture of Vision Encoders (MoVE) has emerged as a powerful approach to enhance the fine-grained visual understanding of multimodal large language models (MLLMs), improving their ability to handle tasks such as complex optical character recognition and scene understanding. Despite these advances, effectively combining diverse encoders and their visual tokens, while also scaling to high-resolution inputs, remains an open challenge. In this work, we conduct a systematic study of fusion designs for MoVE-based MLLMs, highlighting principles for token-level integration across complementary encoders. Our study shows that a lightweight recipe consisting of post-adaptation fusion with independent projectors, tile-level sequence interleaving, and dynamic tiling with global context delivers strong performance on diverse benchmarks. We integrate these principles into a simple and effective architecture that we call LEO. Extensive evaluation on 11 vision-language benchmarks demonstrates that LEO achieves better results on the majority of tasks compared to existing MoVE-based approaches. Furthermore, LEO adapts effectively to the specialized domain of autonomous driving without altering its architecture or training recipe, achieving competitive performance against established baselines and thereby highlighting its ability to generalize. The code is available at https://github.com/Mozhgan91/LEO.

Ensembles of generative large language models (LLMs) are a promising way to compensate for individual model limitations, integrating the strengths of different LLMs. Existing LLM ensemble methods, however, face limitations such as first-token delay and challenges in long-range semantic collaboration between models, Moreover, they typically assume equal voting weights for all models during ensemble, ignoring task-specific performance differences among models. In this work, we propose SpecEM, a training-free, plug-and-play LLM ensemble framework that dynamically adjusts each model's model contribution in real time based on task performance. Inspired by speculative decoding, SpecEM iteratively performs drafting and verification, allowing models to collaborate semantically at the segment level for integrated output. Furthermore, we introduce an online feedback mechanism with multiplicative weight updates, where each model's voting weight is adjusted on-the-fly according to how often it outperforms others during verification stage, ensuring that stronger models exert greater influence during ensembling. Experimental results on five LLM families (ranging from 7B to 72B parameters) and six benchmark datasets, spanning open-domain instruction following, reasoning, commonsense, demonstrate consistent performance improvements compared to state-of-the-art LLM ensemble methods. Our code is available at https://github.com/lvbotenbest/SpecEM.