4157 papers
This paper proposes a time-symmetry-inspired framework for evaluating echo-protected quantum kernel estimation on noisy processors, distinguishing between physical noise reversal and circuit symmetry to establish a rigorous, resource-aware methodology for assessing whether such interventions genuinely improve kernel fidelity and learning performance.
This paper proposes a Human-in-the-Loop framework that combines automated analysis with targeted human review to detect logical vulnerabilities in smart contracts, demonstrating significant improvements in precision and F1-score while reducing false positives on the SmartBugs Curated dataset.
This paper presents a MobileViTv2-based hybrid network enhanced with a normalised cosine-similarity classification head and WGAN-GP synthetic data balancing, which achieves high intra-patient accuracy on the MIT-BIH Arrhythmia Database while highlighting the necessity for future patient-independent validation.
This study demonstrates that Graph Neural Networks, particularly GraphSAGE enhanced with geometric features, significantly outperform traditional thresholding methods in classifying 3D pore connectivity within noisy volumetric images, although improved edge classification does not always guarantee better overall cluster or percolation predictions.
This study demonstrates that for low-resource, imbalanced document error classification, a supervised CNN trained on a cost-effective mix of synthetic and upsampled data outperforms both synthetic-only approaches and large-scale pretrained models, achieving significant gains in accuracy and Cohen's Kappa.
This paper demonstrates that the high accuracy of machine-learning hardware Trojan detectors on standard benchmarks is an artifact of data leakage caused by random trace-level splits, revealing that true cross-Trojan detection performance collapses to chance levels and necessitates per-chip calibration rather than universal pretrained models.
This paper introduces a novel Constraint Programming framework that efficiently mines top-k quantitative sequential patterns under flexible user-defined constraints by leveraging a global constraint for dynamic support filtering and dedicated search strategies to accelerate convergence.
This longitudinal benchmark of five open-weight LLMs across 518 ESG reports reveals that while models reliably summarize narrative policies, they struggle with precise numerical extraction and frequently hallucinate data without explicit reasoning or strict instructions, highlighting a critical need for improved abstention mechanisms in automated sustainability auditing.
This paper demonstrates that while recognizing specific C functions from raw power traces in bare-metal embedded systems is feasible using class-specific autoencoders, the approach faces inherent physical limits due to signal similarities and variations, positioning it as a complementary behavioral signal rather than a standalone security detector.
This paper demonstrates that while length-sorted batch inference can improve processing throughput without sacrificing accuracy in retrieval-augmented generation verification, a simple source-age screening rule is insufficient for detecting factual obsolescence and should be replaced by content-aware temporal checks.