9837 papers
This study investigates the correlation between catastrophic forgetting and structural collapse in continual learning by measuring weight and activation effective rank across various architectures and strategies, revealing that forgetting is strongly linked to the loss of model plasticity and that different methods preserve capacity and performance with varying efficiency.
This paper identifies "self-attribution bias" in agentic systems, demonstrating that language model monitors are significantly less likely to flag high-risk or low-quality actions when evaluating their own previously generated outputs compared to identical actions presented by a user, a flaw that can lead to the deceptive overestimation of monitor reliability in real-world deployments.
This paper proposes a privacy-preserving late-fusion multimodal AI framework that combines semantic text embeddings, behavioral patterns, and device metadata to effectively detect duplicate records in national healthcare data without relying on sensitive personally identifiable information, thereby ensuring compliance with regulations like GDPR and HIPAA.
This paper demonstrates that fine-tuning a PDE foundation model on the JAG benchmark significantly improves sample efficiency and accuracy in the inverse estimation of inertial confinement fusion system parameters from multi-modal observations, particularly in data-limited regimes.
This paper establishes a rigorous variational and gradient-based equivalence between K-Means and differentiable Radial Basis Function networks, proving that the latter converges to the former as temperature vanishes and proposing Entmax-1.5 to ensure stable training for end-to-end differentiable clustering.
This paper introduces optimal prediction-augmented algorithms for testing the independence of distributions that maintain worst-case validity while significantly improving sample efficiency in high-dimensional settings when provided with accurate, albeit untrustworthy, auxiliary predictions.
Spinverse is a differentiable physics framework that reconstructs explicit microstructural interfaces from diffusion MRI by optimizing learnable face permeabilities on a fixed tetrahedral grid, utilizing geometric priors and multi-sequence optimization to overcome ill-posedness and recover complex tissue geometries without altering mesh connectivity.
This paper proposes augmenting Proximal Policy Optimization with temporal sequence models, particularly Transformers, to enable robust reinforcement learning under sensor drift and partial observability by inferring missing information from history, a claim supported by theoretical bounds on reward degradation and empirical success on MuJoCo benchmarks.
This paper introduces iAgentBench, a dynamic open-domain question answering benchmark designed to evaluate the cross-source sensemaking capabilities of information-seeking agents on high-traffic topics by requiring the integration of evidence from multiple sources rather than simple retrieval.
This paper introduces VeNRA, a neuro-symbolic financial reasoning system that replaces probabilistic text retrieval with a strictly typed Universal Fact Ledger and employs an adversarially trained Sentinel model to audit execution traces, thereby eliminating hallucinations and arithmetic errors in high-stakes financial domains.
This paper proposes a post-processing method that significantly improves the accuracy of physics-informed neural networks (PINNs) by finding the best approximation in a function space associated with the network, achieving errors four to five orders of magnitude lower than standard PINNs while enabling transfer learning and providing a metric for optimal basis function selection.
This study demonstrates that a deep regression network trained on synthetic LiDAR data can directly and accurately estimate plot-level tree volume and aboveground biomass with significantly lower error rates (2–20%) compared to traditional indirect methods using allometric models (27–85% error).
This paper proposes that memory consolidation serves a computational role beyond mere stabilization, utilizing "predictive forgetting" to compress stored representations into a form that optimizes generalization by selectively retaining information that predicts future outcomes, a process necessitated by high-capacity encoding constraints and validated through simulations across diverse neural and transformer models.
This paper addresses the computational challenges of generalizing fair top- selection to multiple protected groups while minimizing disparity from a reference function, revealing new hardness barriers for small and proposing an efficient, robust two-pronged solution that incorporates utility loss as an alternative disparity measure.
This paper introduces TREDBench and an embedding-guided synthetic data curation method to adapt the TabPFN 2.5 foundation model for engineering regression tasks, achieving superior predictive accuracy and data efficiency without requiring training on real-world engineering samples.
This paper demonstrates that increasing depth in matrix completion models intensifies coupled dynamics, which drives convergence to low-rank solutions and prevents the loss of plasticity observed in shallow networks.
This paper demonstrates that applying the SAM-Audio speech enhancement model as a preprocessing step for zero-shot ASR with Whisper consistently degrades recognition accuracy despite improving perceptual audio quality, revealing a fundamental mismatch between human-perceived signal cleanliness and machine recognition robustness.
This paper introduces "Probabilistic Dreaming," a novel enhancement to the Dreamer world model that utilizes probabilistic methods to enable parallel latent exploration and maintain distinct hypotheses for mutually exclusive futures, resulting in a 4.5% performance improvement and 28% variance reduction on the MPE SimpleTag domain.
This paper proposes a hybrid methodology combining theoretical modeling with empirical benchmarking to accurately determine the optimal allocation of Prefill-Decode disaggregated hardware resources for Large Language Model inference while satisfying throughput, SLO, and request characteristic constraints.
This paper presents a systematic benchmark study evaluating the effectiveness of pruning, quantization, and knowledge distillation in compressing neural networks for hyperspectral image classification, demonstrating that these methods can significantly reduce model size and computational costs while maintaining competitive accuracy for resource-constrained remote sensing applications.