Aligning large language models (LLMs) with diverse human values is essential for safe and effective deployment, yet existing benchmarks often overlook cultural and demographic variation. We introduce DiverValue-Bench, a population-aware benchmark for evaluating multi-dimensional value alignment across 74 countries/regions. It contains 23,763 quality-controlled instances derived from PRISM user feedback and audited through large-scale human validation, with fine-grained value labels, personalized questions, contrastive reference answers, and rich demographic metadata. Using DiverValue-Bench, we evaluate representative LLMs and reveal substantial geographic and demographic disparities that are masked by aggregate performance. We further show that lightweight preference-based fine-tuning with Low-Rank Adaptation (LoRA) and Direct Preference Optimization (DPO) substantially improves in-domain value alignment while yielding consistent out-of-domain gains. These results highlight the need for population-aware alignment evaluation and demonstrate the utility of DiverValue-Bench as a practical foundation for global alignment, personalized value modeling, and equitable AI development.
论文检索
输入标题、作者或关键词，从 5,485 篇学术成果中精准定位
In subjective tasks, different individuals can have different correct answers for the same input—the ground truth is not fixed but rather determined by each user's personal perspective. Standard foundation models suppress this individual variation by producing population-averaged predictions; conversely, few-shot in-context learning (ICL) struggles to capture user-specific preferences from limited demonstrations. To address this limitation, we propose Population-Guided Bayesian Calibration (PGBC), a test-time personalization method that applies Bayesian inference to foundation model outputs for user-specific alignment. Unlike ICL methods that rely on the model to implicitly infer preferences from examples, PGBC explicitly quantifies each individual's deviation from the population distribution and incorporates this into prediction with minimal feedback (K < 10). Experiments on three subjective benchmark datasets across four foundation models demonstrate that PGBC significantly outperforms zero-shot baselines with a large effect size (Cohen's d = 1.21), and the improvement over ICL approaches is even more pronounced (d = 1.40). Furthermore, user-level distribution analysis reveals that PGBC aligns predictions with individual preferences when zero-shot baselines exhibit substantial bias, while ICL methods increase distributional mismatch. Our results show that PGBC enables practical deployment in cold-start scenarios where minimal user feedback must yield maximal personalization.
Autonomous AI agents on embedded platforms require real-time, risk-aware scheduling under resource and thermal constraints. Classical heuristics struggle with workload irregularity, tabular regressors discard structural information, and model-free reinforcement learning (RL) risks overheating. We introduce GraphPerf-RT, an AI technology achieving deep learning accuracy at heuristic speeds (2-7ms). GraphPerf-RT is, to our knowledge, the first graph-grounded infrastructure unifying task DAG topology, CFG-derived code semantics, and runtime context (per-core DVFS, thermal state, utilization) in a heterogeneous graph with typed edges encoding precedence, placement, and contention. The architecture supports multi-task evidential heads with Normal-Inverse-Gamma uncertainty; we validate on makespan prediction for risk-aware scheduling. Experiments on three ARM platforms (Jetson TX2, Orin NX, RUBIK Pi) achieve R^2 = 0.81 on log-transformed makespan with Spearman rho = 0.95 and conservative uncertainty calibration (PICP = 99.9% at 95% confidence). Integration with four RL methods demonstrates that multi-agent model-based RL with GraphPerf-RT as the world model achieves 66% makespan reduction and 82% energy reduction versus model-free baselines, with zero thermal violations.
The advancement of data privacy regulations has spurred the development of Machine Unlearning (MU), which is designed to remove the influence of sensitive data from a trained model and results in an unlearned model (ULM). Despite rapid progress in MU techniques, their vulnerabilities remain underexplored, which poses risks due to potential leakage of unlearned information. In realistic scenarios, ULMs always need to be incrementally trained with the newly collected data samples, which can lead to the consequences of recalling sensitive information if the new dataset contains similar or even the same unlearned samples. To address this issue, we devise a Durable Unlearning Enhancement (DUE) framework to avoid restoring unwanted sensitive information from incremental training data samples. The DUE framework has three key components that identify sensitive samples and suppress their gradients to update ULMs. Extensive experiments on state-of-the-art MU methods across multiple real-world datasets show that the proposed DUE framework can effectively nullify the recall of sensitive information after MU, and even improve the performance of ULMs. Consequently, our work establishes a new fundamental research direction in safe training against MU vulnerabilities.
Accurate mapping of agricultural land use activity, particularly long-term transition from cropland to pasture and short-term transition between cropland to fallow land, is essential for sustainable water management, drought response, and food-system resilience which directly supports United Nations Sustainable Development Goals (SDG-2 and SDG-8). However, reliable agricultural land use activity mapping is challenging due to spectral ambiguity, temporal irregularities, severe class imbalance, and limited generalization across agricultural regions. In this work, we propose AG-STELLA, a knowledge guided spatio-temporal model that (i) captures temporal changes of agricultural lands using pretrained spatio-temporal transformers; (ii) integrates geospatial context using AlphaEarth embedding; (iii) introduces a temporal transition latent space with temporal consistency constraints; (iv) employs guidance through hydroclimatic consistency; and (v) uses a land use-aware gated decoder to improve robustness across regions. Through experimentation across three water-stressed U.S. states, we show consistent gains over baseline vision and foundation models, achieving up to 27% F1-score improvement for pasture (minority class) and 16% overall. We further show the robustness across heterogeneous regions through cross-state transfer learning, where AG-STELLA consistently outperforms foundation model baselines and achieve up to 82.3% F1-score for fallow land with a 9.6% improvement over the best foundation model. Code: https://github.com/Nibir088/AG-STELLA
The increasing frequency and severity of climate-related disasters have intensified the need for real-time monitoring, early warning, and informed decision-making. Earth Observation (EO), powered by satellite data and Machine Learning (ML), offers powerful tools to meet these challenges. Foundation Models (FMs) have revolutionized EO ML by enabling general-purpose pretraining on large-scale remote sensing datasets. However most existing models rely on high-resolution satellite imagery with low revisit rates, limiting their suitability for fast-evolving phenomena and time-critical emergency response. In this work, we present HighFM, a first cut approach towards a FM for high-temporal-resolution, multispectral EO data. Leveraging over 2 TB of SEVIRI imagery from the Meteosat Second Generation (MSG) platform, we adapt the SatMAE masked autoencoding framework to learn robust spatiotemporal representations. To support real-time monitoring, we enhance the original architecture with fine-grained temporal encodings to capture short-term variability. The pretrained models are then fine-tuned on cloud masking and active fire detection tasks. We benchmark our SEVIRI-pretrained Vision Transformers against traditional baselines and recent geospatial FMs, demonstrating consistent gains across both balanced accuracy and IoU metrics. Our results highlight the potential of temporally dense geostationary data for real-time EO, offering a scalable path toward foundation models for disaster detection and tracking.
Monitoring forest functional diversity is essential to understand ecosystem resilience in the face of rapid environmental change. While existing remote sensing approaches primarily track structural attributes such as canopy density and tree height, functional traits like leaf phenology (evergreen vs. deciduous) and leaf type (broadleaf vs. needleleaf) reveal more direct information about adaptive strategies of tree species. This study presents a scalable machine learning framework for mapping these traits across India at 10 m resolution using Google AlphaEarth Foundations (AEF) embeddings, which capture the complete annual spectral reflectance and radar signatures of the land surface. A key contribution we make is to curate an ML-ready training dataset by combining tree traits information with tree species occurrence data, and to obtain a diverse sample from this data based on spectral time-series to ensure the dataset captures a wide range of phenological dynamics. We then build cross-validation folds to specifically test for spatial generalizability across different eco-regions in India and temporal generalizability across different years, for classifiers learned from the data. Multiple classifiers are evaluated: Random Forest models trained on AEF embedding features achieved the best performance for both classification tasks, outperforming models trained on conventional Sentinel-1 and Sentinel-2 time series while offering seamless deployment in Google Earth Engine. Compared to publicly available land-cover products that encode leaf phenology and leaf type, our model yields significantly higher accuracy while providing outputs at substantially finer spatial resolution. We then observe the outputs of our model over several protected forest areas in India to understand their dynamics over the last 8 years. Our contribution is an analysis-ready open dataset to learn tree traits from remote sensed spectral data, a trained model that is spatially and temporally generalizable, and a demonstration of the insights the model can provide to understand the dynamics of protected forest areas, all of which can be replicated in other areas.
Drought prediction is crucial for disaster mitigation, yet it remains challenging because regional droughts reflect nonlinear interactions among precipitation supply, evaporative demand, and slowly varying land atmosphere states. Although time series foundation models (TSFMs) have shown strong zero shot performance in general forecasting tasks, they can exhibit systematic sensitivity mismatch in the prediction of regional precipitation and temperature anomalies in the Standardized Precipitation Evapotranspiration Index (SPEI). We introduce PhyTTA, a physics informed test time adaptation framework that keeps the TSFM backbone frozen and corrects the predictable part of its one step residual through three components: a dynamic forcing correction, a static tracker of slowly varying background bias, and a conditional confidence gate that applies the correction only in supported regimes. Our analysis shows that, when frozen model residuals contain a component predictable from local forcing changes or slowly varying background bias, correcting this component reduces expected squared error in the population setting. PhyTTA implements this principle through a conservative online ridge EMA correction with regime support gating. Across four foundation backbones on three South Australian locations and TimesFM on three cross regional regions, PhyTTA reduces MSE by up to 19.78%. With the TimesFM backbone it gives consistent 12.3% to 15.1% MSE reductions across the three South Australian locations.
The increasing sensitivity to human aspects in autonomous systems engineering calls for principled approaches to embed ethical and normative concerns into their behaviour. Indeed, recent research has focused on expressing and validating sets of social, legal, ethical, empathetic, and cultural (SLEEC) concerns as rules. However, existing work is limited to rule specification or verification, leaving the problem of semantic-preserving operationalisation of ethical rules in autonomous systems largely unaddressed. For this purpose, we provide an operational solution for ethical-aware autonomous systems, applied to the realm of multi-service robots. Specifically, we devise a principled approach, named CO-SLEEC (Context-Oriented SLEEC), connecting the normative setting of SLEEC rules to context-oriented programming (COP). CO-SLEEC enables runtime adaptation while preserving the semantics of SLEEC rules during robot task execution. It features two reusable Python libraries for (i) parsing SLEEC rules into contextual elements for operationalising them, and (ii) connecting the operational model to the Robot Operating System (ROS), respectively. We evaluate our implementation for correctness, efficiency, and maintainability over multi-service assistive robot scenarios.
In the final "Last-Centimeter" phase of manipulation, where visual occlusion or calibration errors render vision unreliable, robots often suffer high failure rates due to local pose uncertainty and simulation dynamics deviations. To address these issues, this paper proposes the PECHC (Physics-Evolving Cascade Constraint and Human-Correction) algorithm. To rigorously isolate the contribution of tactile feedback in multi-finger coordination, we adopt a decoupled control strategy that focuses on grasp stabilization within the hand's workspace, acting as a fail-safe reflex. The core of our approach is Hybrid Correction Imitation Learning (HCIL), which establishes a "failure-triggered" human-machine mechanism to efficiently resolve the "model gap" via sparse expert corrections. To ensure sample efficiency and baseline performance, we introduce two supporting modules: Cascaded Constraint Scheduling (CCS) addresses the "geometric gap" by enforcing physically plausible behavioral constraints (geometric approach, force closure, and dynamic stability), while Temporal Heterogeneous Distillation (THED) resolves the "physical gap" by enabling implicit system identification from tactile history. Experiments demonstrate that PECHC achieves a 97.3% real-robot success rate on 150 objects from the Visual Dexterity Dataset under fully autonomous testing, where one object is used for one-time HCIL calibration and the remaining 149 objects are evaluated without further intervention. Compared to a standard Sim-to-Real reinforcement learning baseline (Vanilla PPO with Domain Randomization), PECHC delivers a significant performance improvement (+42.8%) and exhibits human-like force modulation capabilities for fragile objects.
Robots that follow natural-language instructions often either plan at a high level using hand-designed interfaces or rely on large end-to-end models that are difficult to deploy for real-time control. We propose TeNet (Text-to-Network), a framework for instantiating compact, task-specific robot policies directly from natural language descriptions. TeNet conditions a hypernetwork on text embeddings produced by a pretrained large language model (LLM) to generate a fully executable policy, which then operates solely on low-dimensional state inputs at high control frequencies. By using the language only once at the policy instantiation time, TeNet inherits the general knowledge and paraphrasing robustness of pretrained LLMs while remaining lightweight and efficient at execution time. To improve generalization, we optionally ground language in behavior during training by aligning text embeddings with demonstrated actions, while requiring no demonstrations at inference time. Experiments on MuJoCo and Meta-World benchmarks show that TeNet produces policies that are orders of magnitude smaller than sequence-based baselines, while achieving strong performance in both multi-task and meta-learning settings and supporting high-frequency control. These results show that text-conditioned hypernetworks offer a practical way to build compact, language-driven controllers for ressource-constrained robot control tasks with real-time requirements.
Robotic perception in unstructured environments remains challenging despite the zero-shot capabilities of foundation models such as SAM. This work attributes performance degradation to non-uniform representation shifts across transformer layers: shallow layers exhibit substantial domain gaps (CKA < 0.5), whereas deep layers transfer effectively (CKA > 0.7). Based on this observation, we propose RepSAM, a representation-guided parameter-efficient fine-tuning (PEFT) framework for adapting foundation models to robotic vision. RepSAM employs a theoretically grounded CKA-guided rank allocation strategy combined with a multi-modal fusion module for robust handling of challenging robotic scenarios, including transparent objects and cluttered scenes. Experimental evaluation across six benchmarks and robotic manipulation tasks demonstrates that RepSAM achieves 97.9% of full fine-tuning performance (89.0% vs. 90.9% mIoU) while reducing trainable parameters by 158× (from 632M to 4.0M). RepSAM outperforms DoRA by 7.9% mIoU with just 4 hours of training on a single A100 GPU (a 96× reduction from full fine-tuning, which takes 384 GPU-hours). These improvements are statistically significant (p<0.01) and translate to a 12.0% absolute improvement in robotic manipulation success rates over the LoRA (RGB) baseline.
The robustness of autonomous vehicles such as drones and Unmanned Surface Vehicles (USV) is crucial when facing unknown and complex marine environments, especially when heteroscedastic observational noise poses significant challenges to sensor-based navigation tasks. Recently, Distributional Reinforcement Learning (DistRL) has shown promising results in some challenging autonomous navigation tasks without prior environmental information. However, these methods overlook situations where noise patterns vary across different environmental conditions, hindering safe navigation and disrupting the learning of value functions. To address the problem, we propose DRIQN to integrate Distributionally Robust Optimization (DRO) with implicit quantile networks to optimize worst-case performance under natural environmental conditions. Leveraging explicit subgroup modeling in the replay buffer, DRIQN incorporates heterogeneous noise sources and target robustness-critical scenarios. Experimental results based on the risk-sensitive environment demonstrate that DRIQN significantly outperforms state-of-the-art meth- ods, achieving +13.51% success rate, -12.28% collision rate and +35.46% for time saving, +27.99% for energy saving, compared with the runner-up.
Offline robotic control requires long-horizon reasoning from fixed datasets while avoiding unsafe extrapolation beyond demonstrated behavior. We propose GRALP, a principled framework that resolves this tension by jointly enforcing support preservation and controllability at the level of temporal abstraction. GRALP adopts a deliberate architectural separation: diffusion is used exclusively as a deterministic action decoder for executing fixed latent skills, while planning and value estimation operate entirely in latent space under conservative constraints. This design enables stable value learning, controllable skill composition, and efficient planning without trajectory-level diffusion sampling at inference. Across unified D4RL benchmarks, GRALP achieves the highest average performance on Navigation, Sequential (Kitchen), and Adroit domains while remaining competitive on locomotion tasks. On contact-rich RoboSuite manipulation with human demonstrations (Lift and Pick-and-Place), GRALP achieves consistently high success rates (over 94%). These results indicate that reliable long-horizon offline control emerges when expressivity is confined to execution and decision-making operates over support-aligned latent abstractions.
Foundation Models for Electronic Health Records (FEMRs) are pretrained on large-scale structured patient data, enabling them to convert longitudinal patient trajectories into generalizable representations for diverse clinical prediction tasks. Despite their effectiveness, FEMRs remain black-box models, raising concerns about bias, interpretability, and clinical trust. To address this, we propose the first token-level explainability approach for FEMRs. We train a Transformer-based surrogate model on input-output pairs from the FEMR across two prediction tasks, approximating its behavior while preserving temporal dynamics. We identify the most influential tokens, providing insights into how FEMRs leverage different aspects of patient history for predictions. To evaluate clinical relevance, we introduce a novel clinical alignment metric that quantifies the correspondence between the surrogate model’s key tokens and clinically validated features. Our results demonstrate that the surrogate closely approximates FEMR predictions and that token-level explanations align well with clinical knowledge, offering a practical framework for interpretable and trustworthy clinical AI.
While Large Language Models (LLMs) offer promise in scientific discovery, leveraging LLMs to drive biomedical research requires the scientific discovery process to be performed in combination with cutting-edge biomedical research and rigorous mechanistic causal chains. As such, both current Retrieval-augmented generation methods lacking causal reasoning capabilities, and the static traditional knowledge graphs failing to reflect evolving scientific knowledge, present obstacles to utilizing LLMs as scientific discovery tools. In response to these ongoing challenges, we present THGAgents. THGAgents utilize collaborative and dynamically updating agents to build a Traceable Causal Knowledge Graph, which serves as the foundation for the evidence-based knowledge structure. Crucially, we employ an LLM-driven heuristic search algorithm to traverse the complex network, balancing both novelty and rigor to deduce strict, evidence-based mechanistic causal chains. Additionally, THGAgents utilize a generator-critic loop to support hypothesis refinement. In experimental benchmarks across both cancer systems and neuroscience, THGAgents achieved up to a 0.80 hit rate in predicting validated scientific discoveries, providing an almost 9.5% increase in hypothesis quality scores versus current state-of-the-art systems, and decreasing the mechanistic hallucination rate to 1.12%. Our code is available at https://github.com/yangCode-res/THGAgents/.
Federated Learning with LoRA fine-tuning offers an efficient and privacy-aware solution for institutions to collaboratively leverage their large datasets to train VLLMs. However, participating institutions often possess heterogeneous computational resources, resulting in imbalanced LoRA ranks, which pose a major challenge for effective collaboration. In addition, real-world applications in domains such as healthcare and transportation frequently suffer from missing modalities due to user mistakes or device failures, which significantly degrade global model performance in federated settings. To the best of our knowledge, no prior work has addressed these two challenges simultaneously in federated VLLMs. To tackle these issues, we propose FediLoRA, a lightweight federated LoRA aggregation framework that effectively mitigates the impact of missing modalities in heterogeneous environment. FediLoRA is explicitly motivated by the observation that simple averaging and structured editing can jointly benefit both global and personalized models. Our approach achieves strong performance across multiple general-domain and medical-domain benchmark datasets. Additional experiments on healthcare data further demonstrate that FediLoRA is well-suited for practical, real-world deployment scenarios. Our code is released at https://github.com/gotobcn8/FediLoRA.
Accurate volumetric medical image segmentation is critical for clinical diagnosis, yet adapting two-dimensional vision foundation models (VFMs) to three-dimensional medical imaging remains challenging. While parameter-efficient fine-tuning (PEFT) methods such as LoRA and adapter-based schemes provide efficient alternatives to full fine-tuning, their geometry-agnostic parameter-space adaptations are insufficient to reconcile discrepancies induced by anisotropic voxel spacing and inter-slice discontinuities. We propose GeoSFLoRA, a geometry-constrained spectral-flow low-rank adaptation framework for efficient 2D-to-3D transfer learning. GeoSFLoRA adapts frozen 2D pretrained vision backbones by operating within a fixed low-rank spectral subspace induced by pretrained linear projections. For each adapted layer, a truncated singular value decomposition is computed once and kept frozen, preserving the original pretrained bases. A lightweight Geometry-Conditioned Encoder (GCE) extracts local volumetric descriptors, which are mapped to token-conditional residuals in the singular-value space, enabling bounded and geometry-aware spectral modulation. GeoSFLoRA consistently improves Dice and HD95 on BraTS20, MSD-Prostate, and MSD-Lung, approaching full fine-tuning performance and demonstrating an effective paradigm for 2D-to-3D medical image segmentation. The code is publicly available at https://github.com/chenbn266/GeoSFLoRA.
Existing physiological foundation models face two limitations: rigid hard-negative sampling indiscriminately repels morphologically similar samples, distorting the natural manifold; and coarse discretization strategies sever the intrinsic continuity of physiological states, inducing precision loss. To address these challenges, we propose DNA-PPG, a novel pre-training framework anchored in Dual Neighborhood Alignment. DNA-PPG integrates the Morphology-Aware Self-Supervised Branch using Time-Frequency Soft Weighting to capture universal signal dynamics shared among diverse subjects, with the Physiological Semantic Alignment Branch that projects physiological indicators into continuous semantic space to explicitly embed precise physiological priors into the representation space. We scale the pre-training to 10.7 million PPG segments from over 8,400 subjects to ensure robust generalization. Extensive evaluations on six downstream benchmarks demonstrate that DNA-PPG significantly outperforms state-of-the-art baselines, achieving an 18% reduction in regression error and an 11.5% improvement in classification performance. These results validate DNA-PPG as a robust, universal feature extractor for diverse photoplethysmography applications.
Foundation models are reshaping medical imaging, yet their application in echocardiography remains limited, hindered by a heavy reliance on private datasets that prevent reproducible comparison. Echocardiography poses unique challenges, including noisy acquisitions, high frame redundancy, and limited diverse public datasets. To address this, we introduce CardioBench, a comprehensive benchmark for echocardiography foundation models. Specifically, CardioBench unifies eight publicly available datasets into a standardized suite spanning four regression and five classification tasks, covering functional, structural, diagnostic, and view recognition endpoints. Leveraging this framework, we evaluate several leading foundation models, including cardiac-specific, biomedical, and general-purpose encoders, under consistent zero-shot, probing, and alignment protocols. Our analysis reveals that while general-purpose encoders transfer well and often close the gap with probing, they struggle significantly with fine-grained distinctions like view classification and subtle pathology recognition. Results indicate that models capturing temporal cardiac dynamics perform best on functional tasks, while retrieval-based approaches generalize more consistently across datasets. By releasing preprocessing, splits, and public evaluation pipelines, CardioBench establishes a reproducible reference point to guide the architectural design of future echocardiography and possibly other medical imaging foundation models.