arXiv:2606.24825v1 Announce Type: new Abstract: Part-of-Speech (POS) tagging is a foundational NLP task underpinning machine translation, information extraction, and syntactic parsing. Despite Marathi being spoken by over 83 million people and ranking among the top twenty most spoken languages worldwide, it remains severely under-resourced in annotated corpora and standardised evaluation benchmarks. Marathi presents unique challenges for computational modelling owing to its rich morphology, relatively free word order, lack of capitalisation conventions, and pervasive code-mixing with Hindi and English. We introduce L3Cube-MahaPOS, a gold-standard POS tagging dataset for Marathi comprising 32,354 manually annotated sentences drawn from news text. Annotation was performed entirely manually by a team of Marathi-proficient annotators following a 16-tag Universal Dependencies-aligned scheme. A structured preprocessing pipeline covering Unicode normalisation, Devanagari-aware tokenisation, and noise filtering ensures label consistency across all splits. We benchmark the dataset across six model families spanning HMM, CRF, BiLSTM, BiLSTM+CharCNN, MuRIL, and the Marathi-specific transformer MahaBERT-v2. The best system achieves 88.67\% token-level accuracy and a macro-F1 of 81.67% over 15 evaluated tag classes. We release the dataset, annotation guidelines, and trained model checkpoints to foster further research in Marathi NLP.
Science Journals
arXiv:2606.24366v1 Announce Type: new Abstract: We present MorfFlex, a morphological dictionary architecture suitable for languages with extensive regularity in both inflection and derivation. As the primary example of MorfFlex in use we introduce MorfFlex CZ, a morphological dictionary of Czech. It is distributed as a simple, unstructured list of triplets, however, its manually maintained, unpublished source files and conversion scripts encode a sophisticated system of inflectional and derivational patterns. These patterns dramatically reduce the otherwise enormous size of the dictionary, which currently contains over 100 million wordforms and more than 1 million lemmas. The MorfFlex CZ dictionary serves as an essential resource for ensuring the consistency of manual morphological annotation in the Prague Dependency Treebanks and underpins state-of-the-art automatic tools such as MorphoDiTa. In this paper, we focus on: (i) presenting an effective method for managing the rich morphological system within the dictionary, and (ii) demonstrating the utility of such a language resource for maintaining annotation consistency in corpora and supporting the development of advanced NLP applications.
Stability Checking of Markov Jump Linear Systems via Probabilistic Temporal Logic (Extended Version)
arXiv:2606.24880v1 Announce Type: new Abstract: Markov jump linear systems (MJLSs) model dynamical phenomena subject to random switching among multiple linear modes, driven by an underlying Markov chain. Classical notions such as mean and mean-square stability characterize the long-term asymptotic behaviour of the first and second moments of an MJLS, but they can be overly conservative or even misleading when only a specific subset of initial conditions is of interest. We tackle this challenge through the lens of model checking, where reasoning about specific sets of initial conditions is intrinsic to the approach. We begin by formalizing probabilistic computation tree logic (PCTL) on MJLSs, enabling the specification of state-based temporal properties for these systems. Building on this foundation, we extend the logic to capture moment-based stability properties relative to a prescribed set of initial states. While we ultimately do not obtain a decision procedure for the entire base logic, the logical extensions can be handled -- albeit with some technical subtleties -- by exploiting linear-algebraic techniques.
arXiv:2606.24883v1 Announce Type: new Abstract: Artificial intelligence (AI) has achieved remarkable success in medical imaging, but it is widely recognized that these models often perform inconsistently across real-world clinical settings. Such inconsistencies occur when patient demographics and imaging protocols vary, for example, in detecting small tumors, analyzing scans from different contrast phases, or evaluating patients of different ages or sexes. To quantify these inconsistencies, we develop a large-scale, open benchmark of 85,355 CT scans that systematically evaluates 12 tumor-detection AI models across tumor size, location, patient subgroup, and imaging protocol. We leverage large language models (LLMs) to extract and organize subgroup information from clinical data, which makes the analysis both scalable and reproducible. Our benchmark reveals that current state-of-the-art AI models, optimized for average accuracy, perform poorly in rare or underrepresented subgroups, such as young, female African Americans. However, collecting sufficient annotated data for these rare cases is often impractical. The benchmark provides a foundation for building more reliable and robust AI models for tumor detection and highlighting the need for rigorous, subgroup-level evaluation in medical imaging and computer vision. Datasets, code
arXiv:2511.08065v1 Announce Type: cross Abstract: Spiking neural networks (SNNs) promise highly energy-efficient computing, but their adoption is hindered by a critical scarcity of event-stream data. This work introduces I2E, an algorithmic framework that resolves this bottleneck by converting static images into high-fidelity event streams. By simulating microsaccadic eye movements with a highly parallelized convolution, I2E achieves a conversion speed over 300x faster than prior methods, uniquely enabling on-the-fly data augmentation for SNN training. The framework's effectiveness is demonstrated on large-scale benchmarks. An SNN trained on the generated I2E-ImageNet dataset achieves a state-of-the-art accuracy of 60.50%. Critically, this work establishes a powerful sim-to-real paradigm where pre-training on synthetic I2E data and fine-tuning on the real-world CIFAR10-DVS dataset yields an unprecedented accuracy of 92.5%. This result validates that synthetic event data can serve as a high-fidelity proxy for real sensor data, bridging a long-standing gap in neuromorphic engineering. By providing a scalable solution to the data problem, I2E offers a foundational toolkit for developing high-performance neuromorphic systems. The open-source algorithm and all generated datasets are provided to accelerate research in the field.
arXiv:2606.23744v1 Announce Type: cross Abstract: Deep learning has become an important tool in computational pathology, enabling automated analysis of histopathological images. While convolutional neural networks (CNNs) have traditionally dominated this field, transformer-based and hybrid architectures have recently demonstrated promising performance. However, comprehensive comparisons of these approaches for colorectal histopathology remain limited. This study evaluated twelve ImageNet-pretrained CNN, transformer, and hybrid architectures using the Kather colorectal histopathology dataset containing 5,000 image tiles from eight tissue classes. All models were trained using a standardized transfer-learning and fine-tuning protocol and assessed using multiple performance metrics, including accuracy, precision, sensitivity, specificity, F1-score, ROC-AUC, Cohen's kappa, and Matthews correlation coefficient. All evaluated models achieved high classification performance, with accuracies ranging from 93.2% to 97.1%. EVA-02 achieved the highest overall performance (97.1% accuracy, 97.0% F1-score), closely followed by ViT-B/16. Among CNNs, ResNet34 and ConvNeXt-Tiny demonstrated highly competitive performance, achieving accuracies of 96.4% and 96.3%, respectively. Transformer architectures generally produced the strongest results across evaluation metrics, although the performance gap between the best transformer and CNN models was relatively small. Per-class analysis showed consistently strong classification performance across all tissue categories, with Complex Stroma representing the most challenging class. Overall, transformer-based architectures achieved the highest predictive performance, whereas modern CNNs provided a favorable balance between accuracy and model complexity. These findings provide a comprehensive benchmark of major deep learning paradigms for colorectal histopathology classification.
arXiv:2511.18037v2 Announce Type: replace Abstract: Hybrid event-frame sensors integrate an Event Vision Sensor (EVS) and an Active Pixel Sensor (APS) within a single chip, combining the high dynamic range and low latency of the EVS with the rich spatial intensity information from the APS. While this tight integration offers compact and temporally precise imaging, the complex circuit architecture introduces nontrivial noise patterns that remain poorly understood and unmodeled. In this work, we present the first unified statistics-based imaging noise model that jointly describes the noise behavior of APS and EVS pixels. Our formulation explicitly incorporates photon shot noise, dark current noise, fixed-pattern noise, and quantization noise, and links EVS noise to illumination level and dark current. Based on this formulation, we further develop a calibration pipeline to estimate noise parameters from real data and provide a detailed analysis of both APS and EVS noise behaviors. Finally, we propose H-ESIM, a statistically grounded simulator that generates RAW frames and events under realistic jointly calibrated noise statistics. Experiments on two hybrid sensors validate our model across multiple imaging tasks, including video frame interpolation and deblurring, demonstrating strong transfer from simulation to real data.
arXiv:2606.23753v1 Announce Type: cross Abstract: CABAC, the entropy coder of H.264/AVC and the basis for HEVC and VVC, decomposes multi-symbol values into bins via a binarization scheme before a binary arithmetic coder. H.264 uses Truncated Unary plus k-th order Exp-Golomb (UEG); alternatives include canonical Huffman and the entropy-conserving binarization (ECB), which provably preserves entropy mapping m-ary data to m-1 binary strings but has not been evaluated inside a production binary arithmetic coder. We integrate ECB into a from-scratch CABAC implementation alongside UEG, single-context Huffman, and a Huffman variant with per-bin-position contexts (HuffmanPos), all sharing one M-coder backend. We benchmark all four on synthetic sources, DCT residuals from a procedural image, and the full 24-image Kodak suite (2,480 round-trip trials, bit-exact verified). On the procedural image, a sparsity-driven crossover at Q=8 lets ECB overtake single-context Huffman, reaching 27 percentage points below at Q=32. On Kodak the crossover shifts below the tested range and ECB beats single-context Huffman at every Q, the gap growing from 0.031 to 0.113 bits per symbol. HuffmanPos, sharing Huffman's codewords but allocating one context per bin position, beats ECB on 12 of 15 source cells and loses by at most 0.56 percentage points on the other three, despite the same per-symbol bin count as single-context Huffman. This isolates the dominant mechanism: at low source entropy the rate gap is driven primarily by context allocation over the bin stream, not the binarization's per-symbol bin count. ECB's rate efficiency costs 7 to 10x in decoder latency on large alphabets, traced to an O(N*m) decoder; we sketch an interleaved single-pass variant that would close this gap. Code, benchmarks, and raw data are open source.
arXiv:2606.23771v1 Announce Type: cross Abstract: Extended Reality (XR) presents a challenging use case for 5G and 6G networks, requiring high data-rates and lowlatency communication to deliver a truly immersive experience. Moreover, in order to seamlessly translate physical actions to the virtual world, accurate gesture recognition and pose estimation are required. Current XR interaction solutions based on handheld controllers and cameras cannot easily capture full-body poses, inhibit the free use of hands, and require good visibility and a clear line of sight. In this work, we propose a multimodal sensing architecture for XR that combines 5G MillimeterWave (mmWave) Integrated sensing and communication (ISAC) and surface electromyography (sEMG) signals. 5G mmWave ISAC cannot only be used to deliver content wirelessly to the Head-mounted display (HMD), but also the same communication signals can be used to derive coarse body-level gestures and poses of the user, to support real-time avatar control. For fine-grained finger-level gestures, our architecture leverages lightweight sEMG sensors that capture forearm muscle activity. To illustrate the need of both modalities, we present evaluations of both sensing technologies. At the body level (5G), our architecture relies on power-per-beam-pair (PPBP), which can be computed from standard beam management or beam sweeping procedures of the 5G NR standard. PPBP-based sensing achieves 82.2$\pm$5.9% average accuracy when evaluated on users not seen during training. For fine-grained finger-level interactions, we show that surface electromyography (sEMG) carries strong discriminative information achieving consistent promising performance across different movement settings. Thus, combining the two modalities enables multi-scale gesture recognition, at the body level via existing 5G signals and finger level via lightweight sEMG sensors, forming a complete XR framework.
arXiv:2606.23925v1 Announce Type: cross Abstract: Kalman filter-based (KF-based) tracking loops are a powerful alternative to traditional phase-locked loops (PLLs) for Global Navigation Satellite Systems (GNSS) signal tracking. The primary advantage of the KF is its ability to incorporate high-fidelity models for receiver dynamics and clock errors, allowing the loop to adapt optimally to signal conditions. However, this theoretical optimality is often compromised in practice by the processing delays inherent in real-time systems with hardware correlators, which existing KF formulations typically neglect. This paper introduces a Modified Kalman filter (mKF) that overcomes this limitation specifically for hardware-based architectures. By reformulating the measurement update to be consistent with the processing delays, the proposed mKF maintains optimality in a practical implementation. We further present a systematic method for tuning both the process noise covariance matrix and the correlation time, based on an analytical expression for the pull-out probability (POP), which is validated through Monte Carlo simulation. The mKF is then validated with a GNSS signal simulator, both by post-processing baseband samples and on a real-time GPS receiver with hardware correlators. A direct equivalence between the mKF and a one-delay Digital PLL (DPLL) is established entirely in the digital domain. At equal noise bandwidth, the mKF matches the DPLL's phase error variance while achieving lower error in the higher-order states. Moreover, the mKF sustains lock at bandwidths inaccessible to the optimal one-delay DPLL under the same dynamic stress, positioning the proposed architecture as a robust and noise-efficient solution for high-dynamic aerospace GNSS applications.
arXiv:2606.24168v1 Announce Type: cross Abstract: Automated diabetic retinopathy (DR) grading from colour fundus photographs can achieve strong predictive performance, but clinical interpretation requires more than an image-level label. It requires understanding how lesion evidence is distributed around retinal vessels and how this evidence relates to quantitative vascular biomarkers. We present a dual-edge spatial-Jacobian image graph for interpretable DR grading. Each fundus image is represented as a graph node with four aligned evidence streams: AutoMorph vessel information ($X_1$), DR-XAI-style lesion evidence maps ($X_2$), a 128-dimensional lesion-based contrastive image embedding ($X_3$), and AutoMorph morphometric biomarkers ($X_4$). The spatial edge branch ($X_{12}$) encodes vessel-lesion geometry, while the Jacobian branch ($X_{34}$) models embedding-biomarker sensitivity. Lightweight two-token attention fuses both edge families into a final image graph. On 2,910 matched non-augmented APTOS images, the full graph achieves 0.8076 accuracy, 0.8312 quadratic weighted kappa, 0.5915 macro-F1, and 0.9330 adjacent-grade accuracy; referable DR reaches 0.9055 accuracy and 0.9711 AUROC. The framework is positioned as an explainable representation-learning tool for lesion-biomarker hypothesis generation, rather than as a deployment-ready clinical classifier. The code is available at https://github.com/Inamullah-Colab/dual-edge-dr-graph-xai.
arXiv:2606.24238v1 Announce Type: cross Abstract: In this work, the ground-state energy of the lithium atom is systematically investigated using both time-independent perturbation theory and the variational method to provide a comprehensive pedagogical analysis of many-body atomic systems. The unperturbed Hamiltonian is initially constructed by neglecting electron-electron interactions, treating the system as three independent hydrogen-like electrons to yield a zeroth-order energy baseline of -275.51 eV. The antisymmetric fermionic nature of the exact wave function is rigorously enforced through the Slater determinant formalism. First-order perturbation theory is applied to evaluate static inter-electronic repulsion using exact Coulomb and exchange integrals, refining the energy state to -192.01 eV. To account for dynamical electronic correlation, second-order perturbation theory is computed numerically for virtual single-electron s-orbital transitions, leading to a total perturbative energy of -196.36 eV. A brief discussion of two-electron excitations is also included to encapsulate further physical realism within the framework. Furthermore, a non-orthogonal two-parameter variational approach is employed to model the shell-specific shielding effect. By optimizing the effective nuclear charges, the variational method establishes a superior upper bound energy of -201.187 eV. The results of both methods are comprehensively contrasted against each other and the reference baseline to provide critical insights into the nature of electron correlation and screening in multi-electron atoms.
arXiv:2606.24393v1 Announce Type: new Abstract: Fibre recruitment plays an important role in tendon and other biological soft tissue mechanics. Due to their large water content, a popular modelling approach for tendons is poroelasticity. Within this framework some tendon studies have included fibres, though none have included crimped fibre recruitment. We present a one dimensional poroelastic model in which the solid skeleton is composed of a soft neo-Hookean background matrix and crimped fibrils which do not bear load (FIB model). As the tissue is stretched, fibrils are straightened and contribute to load bearing. The fibre-reinforced tissue is compared to a tissue with a purely neo-Hookean (NH) skeleton in response to a uniaxial constant applied load (loading) and release of the load (unloading). The system dynamics are governed by a diffusion equation where the diffusion coefficient depends on stiffness. Within tendon parameter ranges, the FIB model is softer than the NH model, and so approaches steady state more slowly during loading. The presence of crimped fibrils allows the tendon to stretch further without excessively straining the fibrils or the NCM, providing a natural protection mechanism for the tendon's structural components to load, in agreement with experiments. During unloading, the FIB model is much slower to relax as the tissue softens due to fibril re-crimping. This asymmetry in loading and unloading manifests as a hysteresis loop in the stress-strain curve averaged over the tendon. The hysteresis is reduced with increasing applied load. The inclusion of fibrils allows for clearer biological interpretation and potential comparison to data. While the stress law employed in this study is bespoke for the application at hand by accounting for crimp and fibril recruitment, other fibril constitutive laws can readily be considered and incorporated into this framework.
arXiv:2606.24402v1 Announce Type: new Abstract: AI security agents increasingly rely on Retrieval-Augmented Generation (RAG) to use external security knowledge for vulnerability analysis and exploit reasoning. This creates a new risk: poisoned write-ups can be operationalized into incorrect exploit behavior. Yet, prior work on RAG poisoning has mostly studied answer corruption in QA settings, much less is known about action-taking security agents. This paper aims to reveal such characteristics with crafted poisons about real-world challenges and AI agents. First, we demonstrate how a crafted single poisoned write-up injected into public-style security knowledge sources which we denote as Poisoned Playbooks, alters the behavior of RAG-based AI security agents. Across 11 CTF challenges, 3 frontier LLM families, 2 model generations, and 11 real-world CVEs, we find that poison adoption is systematic rather than random. To explain this pattern, we introduce the Verification Boundary (VB), a 3-level empirical classification based on what evidence the agent can use to refute a retrieved claim. Finally, we evaluate verification prompting and multi-source retrieval, showing that both help when stronger evidence exists, but weaken under sparse-evidence and zero-day conditions.
arXiv:2606.24483v1 Announce Type: cross Abstract: The deployment of unmanned aerial vehicles (UAV) as open radio units (O-RUs) in 6G cellular systems presents a promising opportunity to achieve scalable and adaptive network coverage. However, optimizing UAV trajectories in dynamic and unfamiliar environments remains a critical challenge, particularly due to the need for extensive retraining in each new scenario. In this paper, we introduce a novel UAV trajectory optimization framework that integrates enhanced continual transfer learning within the O-RAN architecture. The proposed system maintains a library of pre-trained models and employs a model selection mechanism to identify and transfer knowledge from the most relevant environments, minimizing adaptation time and improving efficiency. When no sufficiently similar model is available, a fallback model empowered by continuous refinements ensures baseline performance. The framework leverages real-world city maps and ray tracing techniques to enhance learning reliability and improve trajectory planning. Simulation results demonstrate that the proposed model selection-based transfer learning approach reduces convergence time by 44% to 56% compared to retraining from scratch, and up to 40% compared to traditional transfer learning without model selection.
arXiv:2606.24779v1 Announce Type: cross Abstract: Birth defects are a major cause of fetal loss, neonatal morbidity and long-term disability. In the subset with suspected genetic etiologies, exome and genome sequencing have moved many cases from variant detection to post-sequencing interpretation: clinicians must rank patient-specific candidate variants under incomplete fetal or infant phenotypes and heterogeneous evidence from population genetics, variant-effect prediction, gene-disease validity, phenotype ontologies, cellular and pathway context, protein structure and clinical literature. We present DeepBD, a grounded agentic workflow for variant prioritization and diagnostic interpretation of genetic birth defects. DeepBD organizes the workflow into LLM-assisted case structuring, a pretrained evidence engine, specialist evidence modules and a grounded diagnostic review layer. The evidence engine learns patient-specific variant scores from structured rule evidence, sequence and variant-effect representations and phenotype-conditioned biological context, whereas specialist modules and the agentic layer provide tool-based refinement, candidate-pool review and diagnosis-oriented synthesis from ranked candidates. Developed using an in-house fetal and infant cohort comprising 18,622 cases, DeepBD achieved Recall@1/3/5/10 of 0.658/0.882/0.912/0.929 on an internal held-out solved-case benchmark, outperforming standalone Exomiser, DeepRare and prompted LLM reranking baselines evaluated on Exomiser-derived top-20 candidate variants. Ablation and overlap analyses show that rule evidence, mechanistic context, and specialist refinement provide complementary signals. These findings support a grounded agentic workflow that separates evidence integration, tool-based refinement, and LLM-assisted diagnostic review for retrospective variant prioritization in genetic birth defects.
arXiv:2606.24782v1 Announce Type: cross Abstract: A variational minimum principle for linear elastodynamics of a possibly heterogeneous material without a stored energy function is developed. It involves a change of variables to dual fields, and results in a degenerate elliptic Euler-Lagrange system, even when the primal elastodynamics is hyperbolic. Uniqueness assertions for the dual dynamic and static problems and implications of the degenerate ellipticity are sketched. Some implications pertaining to heterogeneous materials and ones with indefinite elastic moduli are discussed.
arXiv:2111.05385v3 Announce Type: replace Abstract: Obesity is a major health problem, increasing the risk of various major chronic diseases, such as diabetes, cancer, and stroke. While the role of obesity identified by cross-sectional BMI recordings has been heavily studied, the role of BMI trajectories is much less explored. In this study, we use a machine-learning approach to subtype individuals' risk of developing 18 major chronic diseases by using their BMI trajectories extracted from a large and geographically diverse EHR dataset capturing the health status of around two million individuals for a period of six years. We define nine new interpretable and evidence-based variables based on the BMI trajectories to cluster the patients into subgroups using the k-means clustering method. We thoroughly review each cluster's characteristics in terms of demographic, socioeconomic, and physiological measurement variables to specify the distinct properties of the patients in the clusters. In our experiments, the direct relationship of obesity with diabetes, hypertension, Alzheimer's, and dementia has been re-established and distinct clusters with specific characteristics for several of the chronic diseases have been found to be conforming or complementary to the existing body of knowledge.
arXiv:2402.11169v3 Announce Type: replace Abstract: Accurate force fields are essential for reliable molecular simulations. These models are refined against quantum mechanical calculations and experimental measurements, which are subject to random and systematic errors. Bayesian Inference of Conformational Populations (BICePs) is a reweighting algorithm that reconciles simulated ensembles with sparse or noisy observables by sampling the full posterior distribution of conformational populations and experimental uncertainty. In this method, a metric called the BICePs score is used to perform model selection, by calculating the free energy of "turning on" the conformational populations under experimental restraints. This approach, when used with improved likelihood functions to deal with experimental outliers, can be used for force field validation (Raddi et al. 2025). Here, we extend the BICePs approach to perform automated force field refinement while simultaneously sampling the full distribution of uncertainties, using a variational method to minimize the BICePs score. To demonstrate the utility of this method, we refine multiple interaction parameters for a 12-mer HP lattice model using ensemble-averaged distance measurements as restraints. To illustrate the resilience of BICePs in the presence of unknown random and systematic errors, we assess the performance of our algorithm through repeated optimizations and under various extents of experimental error. Our results suggest that variational optimization of the BICePs score is a promising direction for robust and automatic parameterization of molecular potentials.
arXiv:2405.01558v4 Announce Type: replace Abstract: Rendering holograms for holographic displays is often an iterative and computationally costly process. Emerging learned holography methods have alleviated this bottleneck by enabling fast hologram rendering with improved reconstruction quality. However, existing methods still depend on fixed display hardware and scene parameters, requiring retraining for each new configuration. This limits rapid adaptation to different visual needs, including scene brightness, user focus preference, and hardware compatibility. We introduce Configurable Holography, a learned CGH framework in which a single model adapts to diverse display-scene parameters through explicit conditioning, eliminating the need for retraining. As a prototype, we present a configurable structure and derive a family of models that continuously adapt to propagation distance, volume depth, peak brightness, pixel pitch, and wavelength. To further improve efficiency, we incorporate auxiliary monocular depth estimation for depth-aware 3D hologram synthesis from RGB-only inputs and apply knowledge distillation for interactive inference. Our extensive simulation and hardware experiments on three holographic display prototypes with different combinations of configurations show on-par reconstruction quality with existing methods, offering up to 2x speed-up in fp32. Our work represents an initial step toward flexible, general-purpose learned holography systems that can seamlessly adapt across diverse hardware and user-specific visual requirements.
arXiv:2409.11363v2 Announce Type: replace Abstract: AI agents have the potential to aid users on a variety of consequential tasks, including conducting scientific research. To spur the development of useful agents, we need benchmarks that are challenging, but more crucially, directly correspond to real-world tasks of interest. This paper introduces such a benchmark, designed to measure the accuracy of AI agents in tackling a crucial yet surprisingly challenging aspect of scientific research: computational reproducibility. This task, fundamental to the scientific process, involves reproducing the results of a study using the provided code and data. We introduce CORE-Bench (Computational Reproducibility Agent Benchmark), a benchmark consisting of 270 tasks based on 90 scientific papers across three disciplines (computer science, social science, and medicine). Tasks in CORE-Bench consist of three difficulty levels and include both language-only and vision-language tasks. We provide an evaluation system to measure the accuracy of agents in a fast and parallelizable way, saving days of evaluation time for each run compared to a sequential implementation. We evaluated two baseline agents: the general-purpose AutoGPT and a task-specific agent called CORE-Agent. We tested both variants using two underlying language models: GPT-4o and GPT-4o-mini. The best agent achieved an accuracy of 21% on the hardest task, showing the vast scope for improvement in automating routine scientific tasks. Having agents that can reproduce existing work is a necessary step towards building agents that can conduct novel research and could verify and improve the performance of other research agents. We hope that CORE-Bench can improve the state of reproducibility and spur the development of future research agents.
arXiv:2605.06177v2 Announce Type: replace Abstract: Reproducing and comparing deep research agents today is hard: the same backbone evaluated on the same benchmark can report different accuracies across papers because the harness and tool registry differ, and integrating a new model into a comparable evaluation surface costs weeks of model-specific engineering. These are symptoms of a broader reproducibility problem in deep research agent research. Here, we introduce BioMedArena, an open-source toolkit that addresses this reproducibility gap and provides an arena for comparing deep research agents under a shared evaluation environment. BioMedArena decouples six layers of biomedical agent evaluation -- benchmark loading, tool exposure, tool selection, harness mode, context management, and scoring -- and exposes 166 biomedical benchmarks and 75 biomedical tools across 9 functional families. Adding a new model, benchmark, or tool can be accomplished with a few-line provider adapter. Beyond evaluation infrastructure, BioMedArena ships a library of high-quality reference components: 6 agent harnesses (including our proposed Mutual-Evolve) and 6 context-management strategies, any of which can be equipped on any backbone. Equipping these components substantially improves all 12 backbones; on each of 8 representative biomedical benchmarks, the best equipped backbone surpasses prior state-of-the-art (SOTA), by 15.01 percentage points on average. The toolkit, configurations, and per-task traces are available at https://github.com/AI-in-Health/BioMedArena.
arXiv:2501.18502v4 Announce Type: replace Abstract: In this work, we study the problem of distributed mean estimation with 1-bit communication constraints when the variance is unknown. We focus on the setting where each user has access to one iid sample drawn from a distribution belonging to a location-scale family, and is limited to sending just a single bit of information to a central server whose goal is to estimate the mean. We propose simple non-adaptive and adaptive protocols and show that both achieve asymptotic normality. We derive bounds on the asymptotic (in the number of users) Mean Squared Error (MSE) achieved by these protocols. For a class of symmetric log-concave distributions, we derive matching lower bounds for the MSE of adaptive protocols, establishing the optimality of our scheme. Furthermore, we develop a lower bound on the MSE for non-adaptive protocols that applies to any symmetric strictly log-concave distribution, using a refined squared Hellinger distance analysis. Through this, we show that for many common distributions, including a subclass of the generalized Gaussian family, the asymptotic minimax MSE achieved by the best non-adaptive protocol is strictly larger than that achieved by our simple adaptive protocol. We also demonstrate that increasing the number of bits per user can only marginally reduce the asymptotic MSE of adaptive protocols. Our simulation results confirm a positive gap between the adaptive and non-adaptive settings, aligning with the theoretical bounds.
arXiv:2502.01015v5 Announce Type: replace Abstract: Task arithmetic, representing downstream tasks through linear operations on task vectors, has emerged as a simple yet powerful paradigm for transferring knowledge across diverse settings. However, maintaining a large collection of task vectors introduces scalability challenges in both storage and computation. We propose Task Vector Bases, a framework compressing $T$ task vectors into $M < T$ basis vectors while preserving the functionality of task arithmetic. By representing each task vector as a structured linear combination of basis atoms, our approach supports standard operations such as addition, negation, as well as more advanced arithmetic ones. The framework is orthogonal to other efficiency-oriented improvements in task arithmetic and can be used in combination with them. We provide theoretical analysis showing that basis compression retains addition generalization guarantees and enables principled unlearning, with error bounds depending on reconstruction quality. Empirically, our proposed basis construction methods consistently outperform heuristic basis construction baselines and, in some cases, even surpass the performance of full task vector collections across diverse downstream applications while reducing storage and computational requirements. The code is available at https://github.com/uiuctml/TaskVectorBasis.
arXiv:2504.07335v3 Announce Type: replace Abstract: We propose DLTPose, a novel method for 6DoF object pose estimation from RGBD images that combines the accuracy of sparse keypoint methods with the robustness of dense pixel-wise predictions. DLTPose predicts per-pixel radial distances to a set of minimally four keypoints, which are then fed into our novel Direct Linear Transform (DLT) formulation to produce accurate 3D object frame surface estimates, leading to better 6DoF pose estimation. Additionally, we introduce a novel symmetry-aware keypoint ordering approach, designed to handle object symmetries that otherwise cause inconsistencies in keypoint assignments. Previous keypoint-based methods relied on fixed keypoint orderings, which failed to account for the multiple valid configurations exhibited by symmetric objects, which our ordering approach exploits to enhance the model's ability to learn stable keypoint representations. Extensive experiments on the benchmark LINEMOD, Occlusion LINEMOD and YCB-Video datasets show that DLTPose outperforms existing methods, especially for symmetric and occluded objects. The code is available at https://anonymous.4open.science/r/DLTPose_/ .