2787 papers
This paper demonstrates that a unified Graph Neural Network pipeline can successfully learn and generalize multiple fundamental algebraic properties (abelianity, nilpotency, and solvability) from Cayley graphs of finite groups, achieving high accuracy while revealing that different properties require distinct architectural complexities.
This paper presents a significantly simplified and near-optimal self-printing (quine) Turing machine program that not only reproduces its own code from a blank tape but can also print any other program, thereby advancing theoretical models of self-replication and artificial life.
This paper proposes an editable multimodal memory system for long-horizon personalized agents, utilizing a reinforcement learning-based actor-critic manager to dynamically filter, update, and evict heterogeneous evidence while maintaining preference consistency and preventing storage errors, which significantly outperforms baseline methods in recall and memory management across extensive interaction streams.
This study demonstrates that while distilling a value head from a strong Go AI can result in a calibrated but non-discriminative evaluator that breaks MCTS search performance, the search process itself remains a highly effective teacher, as training the policy to mimic MCTS visit counts yields significant strength gains despite the search failing to select better moves during inference.
This study presents a comprehensive failover validation framework that integrates fault injection, state monitoring, and DeepAR-based probabilistic prediction to quantitatively assess and improve cross-region cloud service recovery objectives, successfully reducing median failover time from 31.4 to 12.7 minutes while significantly enhancing data consistency and minimizing secondary failures.
This systematic literature review of 57 studies (2020–2025) evaluates deep learning approaches for CT lesion classification, highlighting the superior performance of hybrid CNN-Transformer architectures and transfer learning while identifying critical gaps in dataset diversity, external validation, and clinical interpretability that currently hinder widespread adoption.
The paper introduces the VQD-CTS Prediction Model, an AI-driven ensemble regression framework that effectively predicts Cost-to-Serve (R² = 0.885) by synthesizing engineering metrics of Velocity, Quality, and Developer Experience, thereby enabling organizations to forecast budgets and optimize investments based on key drivers like code complexity and cycle time.
AutoRE-NS is an explainable neuro-symbolic reinforcement learning framework that significantly enhances industrial requirement quality by integrating semantic analysis, formal validation, and adaptive policy optimization to resolve conflicts, eliminate duplicates, and improve metrics like completeness and atomicity.
This study demonstrates that an on-device edge computing system utilizing a Vision-Language Model (Gemma3-4B) on a Raspberry Pi 5 outperforms a fine-tuned CNN baseline by over 10% in accuracy for real-time disaster detection while enabling natural language response generation with improved energy efficiency and network independence.
This paper presents a comparative study of Variational Autoencoders, Generative Adversarial Networks, Conditional GANs, and CycleGANs across signature synthesis, animal image generation, and face-sketch translation tasks, highlighting their respective performance metrics, architectural strengths, and common training challenges.