3447 papers
The paper presents RSH-SpMM, a fine-grained row-structured hybrid kernel for GPU-based Sparse Matrix-Matrix Multiplication that utilizes adaptive row partitioning, RS-Tile representation, and load-balanced reordering to achieve 1.27x to 6.13x speedups over state-of-the-art methods by effectively handling extreme sparsity irregularity.
FormalRTL is a novel end-to-end multi-agent framework that leverages software reference models as formal specifications to enable scalable, verified, and reliable register-transfer level (RTL) code generation for complex industrial hardware designs.
This paper introduces Kareto, an adaptive multi-objective optimizer that efficiently navigates the complex configuration space of tiered KV cache storage to dynamically balance cost, throughput, and latency, significantly outperforming static strategies in LLM inference services.
This paper proposes extending the ACETONE C code generator, originally limited to sequential code, to support multi-core architectures by formally defining a processor assignment problem and outlining a future implementation of parallel code generation, scheduling heuristics, and synchronization mechanisms.
This paper introduces ChatNeuroSim, an LLM-based agent framework that automates the entire Compute-in-Memory (CIM) accelerator deployment and optimization workflow—including simulation, parameter management, and design space pruning—to significantly reduce the time and effort required for design space exploration.
This paper introduces \texttt{electoral\_sim}, an open-source Python framework that simulates and compares various electoral systems across diverse voter preference distributions using Euclidean distance to the geometric median as a primary performance metric, including a novel theoretical benchmark based on a Boltzmann softmax kernel.
JigsawServe is a novel serving framework that optimizes latency, accuracy, and GPU resource costs for compound inference systems by jointly selecting model variants and spatially partitioning GPUs, achieving up to 11.3x higher throughput and significantly improved resource efficiency compared to prior work.
Granulon is a novel multimodal large language model that leverages a DINOv3-based visual encoder enhanced with a text-conditioned granularity controller and adaptive token aggregation to dynamically unify pixel-level perception with coarse-grained semantics, significantly improving accuracy and reducing hallucinations compared to existing approaches.
This paper introduces the Elastic Time Warping algorithm, which solves the time series matching problem in arbitrary metric spaces using a Hellinger kernel stretching penalty with cubic computational complexity.
This paper presents a representation-native framework for explainable 3D Gaussian Splatting watermarking that utilizes a Trio-Experts module and a Safety and Budget Aware Gate to optimize carrier selection and visual compensation, achieving superior robustness and fidelity while providing auditable insights into the watermarking process.
This study investigates how age-related differences in sensitivity to deictic gaze affect social perception during human-robot interaction, aiming to inform the design of adaptive non-verbal cues for social robots assisting older adults.
The paper introduces VisionCreator-R1, a native visual-generation agent enhanced with explicit reflection mechanisms and trained via a Reflection-Plan Co-Optimization (RPCO) methodology that addresses credit assignment challenges to outperform state-of-the-art models on both single and multi-image generation benchmarks.
This paper addresses the lack of standardized benchmarks and datasets in embodied healthcare by introducing MedMassage-12K, a large-scale multimodal acupoint massage dataset, and proposing HMR-1, a hierarchical framework that leverages vision-language models for high-level acupoint grounding and low-level trajectory control to enable robust robotic massage therapy.
This paper reflects on the barriers to interdisciplinary collaboration in HCI and aging research by analyzing the disconnect between technological design and older adults' needs, while sharing the authors' personal journey of developing empathy through volunteering in a senior living community.
This study demonstrates that in human-robot interaction, different failure types impact perceived reliability differently—with mistakes being less damaging than slips or lapses—and that trust can be effectively recovered through subsequent successful executions without the need for explicit social repair actions.
This paper resolves the open question regarding the parameterized complexity of QBF with few existential variables by proving that the previously known double-exponential dependence on the number of existential variables is optimal under the ETH for general CNFs, while demonstrating that the problem becomes significantly more tractable for formulas with only two quantifier blocks.
This paper proposes a cost-effective, computer vision-based smart parking system that utilizes YOLOv8 for vehicle detection and inverse perspective mapping to merge multi-camera views into a simulated 3D environment for guiding users to vacant spots.
The paper proposes HeteroFedSyn, the first differentially private framework for synthesizing tabular data in horizontal federated settings, which achieves utility comparable to centralized methods by introducing noise-efficient dependency metrics, unbiased noise correction, and adaptive selection strategies to handle heterogeneous data distributions.
This paper presents NaviNote, a voice-based system that combines high-precision visual localization with an agentic architecture to enable blind and low vision users to author in-situ spatial annotations and navigate unfamiliar environments with greater accuracy.
This study reveals that the impact of LLMs on critical thinking is fundamentally shaped by time constraints, where early or continuous access enhances performance under time pressure but impairs it when sufficient time is available, whereas delayed or no access yields the opposite effect.