As AI‑based decision‑support systems become increasingly widespread, methods aimed at improving the performance of human-AI teams are gaining attention. In recent years, explainable artificial intelligence (XAI) has received growing interest, as it provides methods to make the behavior of machine learning models more transparent and can help to identify errors and flaws, which is particularly important in safety critical domains such as medicine and law. However, explanations themselves can be misleading, inconsistent, or incorrect, which makes it essential to raise awareness of the possibilities and limitations of these methods. We introduce Lucyde, a web‑based demonstrator designed to help users explore, compare, and better understand XAI methods across different datasets, models, and configurations. Lucyde provides a curated collection of explanation techniques, enables side‑by‑side comparison of methods, and offers easy‑to‑understand supplementary information for different user groups. It also illustrates interactive machine learning workflows by allowing users to correct model outputs or explanations. Lucyde thereby fosters informed and reflective engagement with AI systems and their explanations.
论文检索
输入标题、作者或关键词，从 3,443 篇学术成果中精准定位
The increasing use of neural networks (NNs) in high-stakes decision-making requires rigorous analysis to ensure safety, fairness, and explainability. Formal verification tools for neural networks typically focus on determining the satisfiability of properties such as safety or fairness. However, many fairness and explainability tasks go beyond satisfiability and instead rely on arbitrary optimization objectives. To address such problems, neural networks are often encoded as mixed-integer linear optimization (MILO) problems with linear objectives. In practice, these encodings are usually implemented in an ad-hoc manner, limiting comparability across works, reducing transparency, and increasing implementation effort. We address this gap by introducing eNNcode, a user-friendly PyPI library that converts any piecewise-linear neural network in Open Neural Network Exchange (ONNX) format into a MILO instance. The library supports arbitrary constraints on input and output nodes, as well as user-defined optimization objectives. Our experiments show that eNNcode achieves performance comparable to existing libraries, despite its simplicity and ease of use. Overall, eNNcode facilitates reproducible and standardized optimization-based analysis of neural networks.
While we have witnessed extraordinary progress across various subfields of AI, the deployment of these systems in safety-critical and mission-critical domains has lagged behind. A key requirement for such deployments is the availability of AI systems capable of generating optimal behavior that can be formally verified and effectively used by non-AI experts from diverse backgrounds. While significant advances have been made toward building such systems, we still lack comprehensive formal frameworks to model and analyze the complex dynamics of human-AI interaction. In this talk, I will introduce the Human-Aware AI framework, a multi-agent planning framework specifically designed to support and reason about human-AI interaction. I will show how this framework provides novel solutions to challenges such as explainability, value alignment, and proactive assistance, and demonstrate its application in domains including intelligent tutoring systems, cybersecurity, and robotics.
Large language models (LLMs) require substantial investments and are increasingly deployed in high-stakes domains, making it critical to protect LLM-related assets and to trace their provenance. Identity technologies such as fingerprinting and watermarking address these needs by enabling ownership verification and attribution, and have rapidly emerged as an active research focus. However, existing techniques lack a systematic organisation, leading to two key issues, terminological confusion and isolated research lines, that have hindered the development of this research field. To this end, we present a comprehensive review of LLM identity techniques, focusing on fingerprinting and watermarking across the LLM lifecycle, including datasets, models, and generated content. We make three primary contributions. First, we introduce implicit identity (Implicit-ID for short) as a unifying abstraction and distinguish fingerprinting from watermarking. Second, we propose a lifecycle-based taxonomy that organises techniques by asset type and verification role, aligning each with asset protection or provenance. Third, we establish an evaluation framework around three objectives---identifiability, robustness, and deployability. Together, these contributions structure the landscape of LLM identity techniques, clarify terminology, and highlight directions toward secure deployment.
Large Language Models (LLMs) have become central to modern AI, with interpretability playing an increasingly important role in investigating the opaque and complex mechanisms encoded within billions of parameters and ensuring trustworthy deployment. However, descriptive interpretability approaches for LLMs remain largely post-hoc, explaining model behavior without actionable guidance for model improvement, thereby limiting their practical utility. Recent work has therefore shifted toward actionable LLM interpretability, emphasizing methods that connect internal mechanisms to interventions and model improvement rather than explanation alone. This survey reviews LLM interpretability through the lens of actionability, presenting a taxonomy of attributional and mechanistic approaches, along with emerging methods tailored to vision–language models (VLMs). We further examine how actionable interpretability supports downstream objectives such as hallucination mitigation, knowledge editing, fairness, safety, capability and efficiency. By positioning actionable interpretability as a pathway for better-guided LLM design and practice, this survey outlines key challenges and future directions toward trustworthy and controllable foundation models.
Reinforcement learning (RL) is increasingly applied in complex, safety-critical domains, yet the lack of rigorous behavioral guarantees for neural network-based policies remains a major barrier to deployment. Recent advances in policy expressiveness and scale have intensified this challenge, leading to a rapidly growing but conceptually fragmented body of work on RL policy verification. This survey provides a unifying perspective on RL verification methods. We introduce a taxonomy that clarifies relationships among existing approaches along three axes: verification paradigm (formal versus probabilistic), temporal scope (step-wise versus multi-step), and guarantees strength. Beyond taxonomy, we unify underlying theoretical foundations, make implicit assumptions and limitations explicit, and identify emerging directions.
Deep neural networks are vulnerable to adversarial examples, posing serious security risks. While GAN-based attacks such as AdvGAN enable fast adversarial example generation, they typically operate in a single step, limiting attack effectiveness and transferability. In this work, we propose PAR-AdvGAN, a progressive auto-regressive GAN framework that iteratively refines adversarial perturbations. By conditioning each iteration on the previous output, PAR-AdvGAN produces stronger, more transferable attacks while maintaining low visual distortion and high generation speed. Experiments on multiple models and datasets demonstrate that PAR-AdvGAN significantly outperforms baseline methods in attack success rate, achieving up to 335.5 frames per second, highlighting its practical efficiency for black-box attack scenarios.
Graph Neural Networks (GNNs) have demonstrated remarkable proficiency in modeling data with graph structures, yet recent research reveals their susceptibility to adversarial attacks. Traditional attack methodologies, which rely on manipulating the original graph or adding links to artificially created nodes, often prove impractical in real-world settings. This paper introduces a novel adversarial scenario involving the injection of an isolated subgraph to deceive both the link recommender and the node classifier within a GNN system. Specifically, the link recommender is mislead to propose links between targeted victim nodes and the subgraph, encouraging users to unintentionally establish connections and that would degrade the node classification accuracy, thereby facilitating a successful attack. To address this, we present the LiSA framework, which employs a dual surrogate model and bi-level optimization to simultaneously meet two adversarial objectives. Extensive experiments on real-world datasets demonstrate the effectiveness of our method.
Chlor-alkali production is a large-scale industrial process whose operating conditions and equipment states evolve over time. Its process optimization requires ongoing trade-offs among conflicting objectives such as product yield, energy consumption, and equipment life. Existing optimization approaches are typically static and must be re-optimized after environmental changes, limiting their real-world applicability. In this work, we model the problem as a dynamic multi-objective sequential decision-making problem that continuously tracks a time-varying Pareto set under changing conditions. We propose MORL-CA, a multi-objective reinforcement learning framework that integrates offline pretraining on historical data with constrained online policy refinement. MORL-CA introduces a state-aware adaptive objective weighting mechanism within a multi-critic actor-critic architecture, enabling localized Pareto-improving policy updates while satisfying operational and safety constraints. Extensive experiments in an environment conducted from real chlor-alkali data demonstrate that MORL-CA achieves superior Pareto solution quality and smoother adaptation to dynamics compared with state-of-the-art multi-objective optimizers and MORL baselines.
Flood models inform strategic disaster management by simulating the spatiotemporal hydrodynamics of flooding. While physics-based numerical flood models are accurate, their substantial computational cost limits their use in operational settings where rapid predictions are essential. Models designed with graph neural networks (GNNs) provide both speed and accuracy while having the ability to process unstructured spatial domains. Given its flexible input and architecture, GNNs can be leveraged alongside physics-informed techniques with ease, significantly improving interpretability and generalizability. We introduce a novel flood GNN architecture, DUALFloodGNN, which embeds physical constraints at both global and local scales through explicit loss terms. The model jointly predicts water volume at nodes and flow along edges through a shared message-passing framework. To improve performance for autoregressive inference, model training is conducted with a multi-step loss enhanced with dynamic curriculum learning. Compared with standard GNN architectures and state-of-the-art GNN flood models, DUALFloodGNN achieves substantial improvements in predicting multiple hydrologic variables (e.g., water volume, flow, and depth) while maintaining high computational efficiency. The model is open sourced at https://github.com/acostacos/dual_flood_gnn. The dataset is open sourced at https://doi.org/10.25910/9xav-0s86.
Reinforcement learning (RL) policies often outperform classical controllers in simulation, yet rarely reach production in safety-critical processes. The barrier is that there is no principled way to answer “Is this policy safe to deploy?” with statistical guarantees. We introduce DEPLOY-RL, a post-training certification framework built on one key insight: deployment certification requires discovering where failures occur (boundary discovery), not measuring how much everywhere (uniform reconstruction). Our contributions: (1) a contract-coupled acquisition function that concentrates sampling on certification-critical boundaries, achieving ≈ 2× sample efficiency with a semi-empirical ambiguity reduction bound (domain-calibrated convergence guarantee); (2) conformal risk control providing finite-sample false-go guarantees (≤ α) under explicit deployment contracts; (3) a three-way decision framework (Deploy/No-Deploy/Abstain) with fail-safe PID/MPC fallback. In simulations on papermaking (industrial digital twin) and Tennessee Eastman (public benchmark), DEPLOY-RL achieves 4.4% false-go rate (vs. 8.6% for the best baseline) while retaining 88.6% policy coverage—the only method achieving <5% false-go with >85% coverage among 14 baselines under our evaluation protocol.
Dynamically Reconfigurable Battery (DRB) systems employ power electronic switches to create dynamic topologies. They enable effective management of cell difference through real-time adjustment of cell connections. However, existing DRB control methods struggle to learn effective strategies due to sparse rewards, which arise from blind exploration in large topological action spaces and complex operational constraints. This leads to insufficient policy learning, making safety and balancing performance difficult to ensure in practical applications. To this end, we propose a Subgraph-Augmented Hierarchical Reinforcement Learning (SAHRL) framework. By combining hierarchical policies with topological structural knowledge, SAHRL effectively accelerates policy exploration and mitigates reward sparsity. Specifically, the high-level policy determines the strategic direction, while the subgraph-augmented low-level policy refines actions to meet operational constraints. The topological structural knowledge, extracted in the form of subgraphs and incorporated as an inductive bias, helps the agent focus on meaningful action patterns and reduce invalid exploration in the large action space. Extensive simulations and real-world experiments show that SAHRL achieves safe and efficient balancing. Notably, it increases the energy release by 10.56% compared to conventional methods in real-world applications.
Anterior cruciate ligament (ACL) injury is one of the most common and serious injuries in sports, particularly among recreational skiers. Research shows that structured technique awareness and continuous feedback can significantly reduce the risk of such injuries, yet access to professional instructors is limited to wealthy athletes who can afford continuous private coaching, creating a harmful inequity in injury prevention. This gap can be mitigated by automating the real-time analysis of skiing techniques available to the wider recreational skiing community. The approach relies exclusively on inertial sensors embedded in standard smartphones, eliminating the need for specialized equipment and enabling broad social scalability. To support immediate feedback, the system operates causally, producing predictions based solely on past observations. The work is conducted in cooperation with professional ski instructors, ensuring that problem formulation, data annotation, and result evaluation reflect real-world coaching practices and injury prevention needs. The model is evaluated using Leave-One-Subject-Out validation on a public, in-the-wild dataset, demonstrating robust generalization across skiers, achieving an average directional accuracy of 89.8%, while maintaining extremely low inference latency suitable for on-device mobile deployment. This work outlines a practical pathway to democratizing injury prevention in recreational sports.
Effective flood risk management relies on accurate forecasting, yet the ”black box” nature of state-of-the-art Deep Learning models creates a barrier to trust and accountability in high-stakes public safety decisions. While existing Explainable AI (XAI) methods offer local attributions, they fail to provide the verifiable, operationally meaningful causal narratives required by disaster response authorities. To address this societal challenge, we propose Context-Aware Concept Distillation (CACD), a framework developed in collaboration with domain experts to distill opaque LSTMs into interpretable, hydrology-aware surrogate models. We introduce an unsupervised pipeline to discover a ”Hydrological Language” and a Residual Hypernetwork that dynamically modulates these concepts based on static basin characteristics. Evaluated on 5,203 basins globally, our model achieves high fidelity (Median NSE 0.70), significantly outperforming black-box baselines (e.g., Multi Layer Perceptrons) on unseen future data. By demonstrating that human-interpretable concepts are sufficient to reconstruct flood dynamics, this work balances AI accuracy with the transparency required for responsible environmental decision-making.
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.
Vision-language-action models (VLAs) often achieve high performance on demonstrated tasks but struggle significantly when required to extrapolate, recombining skills used in different tasks in novel ways. For instance, VLAs might successfully put the cream cheese in the bowl and put the bowl on top of the cabinet, yet still fail to put the cream cheese on top of the cabinet. This motivates us to investigate whether VLAs merely overfit to demonstrated tasks or still hold the potential to extrapolate. Our study uses text latent as the ingredient; it is a task-specific vector derived from the models’ hidden states. It thus encodes semantics necessary for completing a task and can be used to reconstruct the associated task behavior by writing it to the model’s residual stream. Furthermore, we find that skills used in distinct tasks can be combined to produce novel behaviors by blending their respective text latent. Applying this to π0, we increase its success rate from 9% to 83% on the proposed libero-ood benchmark, which features 20 tasks extrapolated from standard LIBERO tasks. This reveals that the skill representations encoded in text-latent are individual yet composable, while π0 fails to autonomously combine these representations for extrapolation. This also validates the design of libero-ood; it comprises tasks that the model fails, yet should be able to complete. We then tested other VLAs on libero-ood, and none of them achieved a success rate higher than 21%. Further analysis reveals VLAs share a common pattern to exhibit spatial overfitting, associating object names with where the object is spatially located in the demonstrated scene rather than achieving true object and goal understanding.
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.
Safe and precise maneuvering of quadrotor unmanned aerial vehicles (UAVs) in high-speed wind environments remains a critical challenge. Wind disturbances are nonlinear, time-varying, and difficult to model, causing traditional controllers to struggle with perception and compensation, especially under unseen wind distributions. To address these limitations, we introduce WA-TD3, a data-driven control framework that enables real-time wind disturbance perception and adaptive compensation without dedicated wind sensors. WA-TD3 employs a deep residual network to extract wind characteristics from temporal patterns in state deviations, forming a dynamics residual-driven perception mechanism that implicitly models and compensates for unknown winds. This residual is integrated into a perception-augmented reinforcement learning architecture, providing the policy with enhanced state information for proactive disturbance-aware control. Extensive experiments on complex trajectories under varying wind intensities demonstrate that WA-TD3 consistently outperforms state-of-the-art methods, achieving over 62% improvement in tracking accuracy under strong winds.
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.
Large vision-language models (VLMs) demonstrate strong performance in medical image understanding, but frequently generate clinically plausible yet incorrect statements, raising significant safety concerns. Existing medical hallucination benchmarks primarily focus on 2D imaging with one-shot diagnostic questions, offering limited insight into whether predictions are grounded in correct localization and abnormality identification, allowing critical reasoning errors to remain hidden behind seemingly correct diagnoses. We introduce Med-StepBench, the first large-scale benchmark for step-wise hallucination detection in 3D oncological PET/CT, comprising over 12,000 images and more than 1,000,000 image–statement pairs across volumetric and multi-view 2D data, which decomposes clinical reasoning into four expert-designed diagnostic stages. Using clinician-verified annotations, we perform the first step-level evaluation of general-purpose and medical VLMs, revealing systematic failure modes obscured by aggregate accuracy metrics. Furthermore, we show that current VLMs are highly susceptible to adversarial yet clinically plausible intermediate explanations, which significantly amplify hallucinations despite contradictory visual evidence. Together, our findings highlight fundamental limitations in grounding multi-step clinical reasoning and establish Med-StepBench as a rigorous benchmark for developing safer and more reliable medical VLMs.