7103 papers
This paper introduces CLAIRE, a hybrid deep learning framework that combines unsupervised latent representation learning with supervised classification and game-theoretic interpretability to achieve robust, explainable fault detection in high-dimensional industrial environments.
This paper introduces ALFCG, the first adaptive, projection-free framework for stochastic composite nonconvex optimization that eliminates the need for global smoothness constants or line search by using self-normalized accumulators to estimate local smoothness, achieving optimal iteration complexity up to logarithmic factors while outperforming state-of-the-art baselines.
This paper extends kinetic-based regularization to accurately estimate spatial derivatives from noisy data through explicit and implicit schemes, demonstrating its effectiveness in enabling stable shock capture for 1D hyperbolic PDEs while preserving conservation laws.
This paper proposes a schema-gated agentic AI architecture that resolves the trade-off between conversational flexibility and execution determinism in scientific workflows by enforcing machine-checkable specifications as mandatory execution boundaries, a solution validated through multi-model LLM scoring of 20 existing systems.
The paper proposes R4T, a three-stage framework that leverages reinforcement learning to synthesize objective-aligned training data for a lightweight diffusion retriever, enabling efficient, high-quality set-valued retrieval that optimizes complex properties like diversity and coverage while significantly reducing inference latency compared to RL-based baselines.
This paper addresses the challenges of XL-MIMO radiomap prediction in the upper 6 GHz band by introducing a comprehensive multi-configuration dataset, a systematic benchmark framework, and a novel physics-informed "beam map" approach that significantly improves generalization to unseen array configurations and environments by decoupling deterministic radiation patterns from learned multipath propagation.
This paper proposes M-CMAB, a multi-adapter-enhanced contextual multi-armed bandit framework that optimizes online multi-modal large language model inference scheduling under heterogeneous, time-varying constraints by combining a frozen-backbone predictor with lightweight adapters, a primal-dual constraint handler, and a two-phase scheduler to maximize rewards while respecting irreversible multi-dimensional budgets.
This paper proposes a reference architecture for reinforcement learning frameworks by analyzing 18 state-of-the-practice implementations through grounded theory to establish a common basis for comparison, evaluation, and integration.
The paper proposes CLoPA, a continual low-parameter adaptation strategy that efficiently tunes a small fraction of the nnInteractive model on incoming annotation data, rapidly achieving expert-level performance across diverse medical imaging tasks without requiring new parameters or altering the inference pipeline.
This paper presents a framework for the certified and accurate computation of deep neural network function space norms (including , , and ) by combining interval arithmetic, adaptive refinement, and quadrature to derive guaranteed global bounds from local certificates, thereby enabling reliable error control for PDE applications like PINNs.
This paper introduces a Correlation-Complexity Map, comprising Quantum Correlation-Likeness and Classical Correlation-Complexity indicators, to identify real-world data distributions suitable for IQP-based quantum generative models, demonstrating that such models can achieve competitive performance with fewer resources on complex turbulence data when guided by this diagnostic framework.
This paper demonstrates that in quantum Gaussian dynamics, reversing a noising process using a fixed-diffusion Wigner-score drift generally violates complete positivity unless specific conditions are met, thereby necessitating the injection of additional diffusion to ensure physical validity and incurring a fundamental cost in fidelity.
The paper introduces NOBLE, a pretraining architecture that permanently augments transformer linear layers with learnable nonlinear low-rank branches (specifically using CosNet activation), achieving significant training efficiency and speedups across various models with minimal parameter and time overhead, though its benefits may be hindered by certain stochastic data augmentations.
COLD-Steer is a training-free framework that steers large language models by approximating the representational changes of gradient descent on in-context examples, achieving up to 95% steering effectiveness with 50 times fewer samples than existing methods.
This paper introduces a semantics-aware caching approach that significantly reduces the runtime of concept learning by efficiently managing instance retrieval across diverse symbolic and neuro-symbolic reasoners, achieving speedups of up to an order of magnitude.
This paper challenges the assumption that multimodal backdoor attacks are synergistic by revealing a "winner-takes-all" phenomenon called Backdoor Modality Collapse, where attacks degenerate to rely predominantly on a single modality, and introduces new metrics to quantify this vulnerability.
This paper presents an artificial intelligence system trained on over 45,000 ultrasound images that achieves diagnostic accuracy comparable to senior radiologists for fetal orofacial clefts, significantly enhances junior radiologists' performance when used as a copilot, and accelerates clinical expertise development for rare conditions.
This paper demonstrates how transformer models can serve as computationally efficient surrogates to accurately predict atomistic transition pathways in nano-clusters, enabling the generation of diverse microstates while overcoming the high costs of conventional simulation techniques.
This paper introduces a novel interpretability method for large hierarchical probabilistic time-series forecasting that addresses structural and uncertainty challenges, successfully explaining state-of-the-art industrial models to enhance stakeholder trust and decision-making through real-world case studies and semi-synthetic evaluations.
This paper introduces CODEC, a method that utilizes sparse autoencoders to decompose neural network computations into sparse, causal motifs of hidden-neuron contributions, thereby revealing how nonlinear processes evolve across layers and enabling greater interpretability and control of both artificial and biological neural systems.