4157 papers
The paper introduces Warp-Geo, a differentiable, GPU-accelerated framework that combines uniform grids with implicit Poisson differentiation to enable efficient, frequent geometry updates for gradient-based shape optimization and dynamic boundary simulations.
This paper proposes a synergistic design framework that integrates C/K theory with frame semantic hypergraphs and topological diagrams to navigate radical transitions in generative AI, addressing the limitations of current models in handling conceptual and data shifts while balancing representational correctness with performative efficacy.
This paper proposes a multi-task deep learning framework that integrates complementary spectro-temporal features with a hybrid Bi-LSTM/Transformer architecture to simultaneously achieve high-accuracy deepfake speech detection and known spoofing attack attribution, demonstrating robust performance on the ASVspoof 2019 dataset while highlighting the need for future optimization against severe acoustic noise.
The paper proposes TAHAN, a novel Type-Aware Heterogeneous Graph Attention Network that enhances node representation learning by organizing multi-level neighbors into type-specific groups and applying a two-stage attention mechanism to dynamically capture and fuse type-dependent contextual signals, thereby outperforming existing heterogeneous graph neural networks in various tasks.
This paper introduces the first shared benchmark for automatic determinative detection in ancient Egyptian hieroglyphs and proposes a robust multi-agent LLM framework that combines semantic and structural reasoning with deterministic constraints to achieve high accuracy and transparent interpretability in this low-resourced domain.
This paper presents an iterated local search algorithm with automatically configured parameters and a public benchmark of 324 instances to solve the Book Scanning Problem, achieving superior solution quality compared to existing methods while proving optimality for 21 cases.
This paper introduces Seasonality-Aware Dual-Memory Pricing (SA-DMP), a learning algorithm that dynamically balances long-term seasonal patterns and short-term market shifts to optimize electric vehicle charging prices under non-stationary demand, demonstrating statistically significant regret reduction compared to fixed-weight baselines while identifying specific operational boundaries where adaptive pricing outperforms traditional time-of-use tariffs.
The paper proposes DFARNet, a dual-attention ResNet architecture that combines learnable frequency and spatial attention mechanisms to effectively distinguish snow particles from natural edges, achieving a PSNR of 31.6 dB on the CSD dataset while remaining suitable for resource-constrained deployment.
This paper introduces "molecular traps"—deliberately designed molecules that exploit common algorithmic weaknesses—to systematically expose significant failures in current computer-aided synthesis planning tools regarding chemoselectivity, regioselectivity, and steric hindrance, thereby offering a critical stress-testing framework for developing more practical and reliable synthesis algorithms.
This paper presents a self-healing framework for small language models that combines syntax repair, schema-guided conversion, and validator feedback to significantly improve JSON validity from 16.7% to 78.1% without modifying model weights, while revealing that such syntactic corrections do not guarantee semantic accuracy.