AI论文速递 2026年05月21日（HuggingFace Daily Papers）¶

数据来源：https://huggingface.co/papers 采集时间：2026-05-21

📌 重点关注¶

MobileVLN: On-Device Learning for Vision-Language Navigation | arXiv — 【重点关注】 Vision-Language Navigation (VLN) enables agents to navigate and interact with environ... 💡 端侧VLN训练，移动端AI应用落地新突破
EfficientAgent: Accelerating Large Language Models through Hierarchical Task Decomposition | arxiv.org — 【重点关注】 Large language models (LLMs) are increasingly used for complex reasoning tasks... 💡 分层任务分解，Agent效率提升关键技术
EvoLM: Evolutionary Prompt Optimization for Enhanced LLM Performance | arxiv.org — 【重点关注】 Prompt engineering remains a critical bottleneck for maximizing the performance... 💡 进化式Prompt优化，AI工具开发新范式

MemoryOpt: Compressed Long-Term Memory for Efficient Agent Deployment | arxiv.org — Large language models deployed as agents require efficient memory management to handle...
EdgeAgent: Lightweight On-Device AI for Mobile Computing | arxiv.org — The proliferation of mobile devices has created an unprecedented demand for on-device...
ToolChain: Automated Tool Discovery and Integration for LLM Agents | arxiv.org — Large language models (LLMs) increasingly benefit from external tools to extend their...
SparseAttention: Reducing Computational Overhead in Vision-Language Models | arxiv.org — Vision-language models (VLMs) have demonstrated remarkable capabilities in multimodal...
ReasonChain: Improving Multi-Step Reasoning through Chain-of-Thought Optimization | arxiv.org — Large language models (LLMs) have shown impressive performance on single-step reasoning...
VisionGPT: Enhanced Visual Understanding through Multimodal Context Integration | arxiv.org — Vision-language models (VLMs) have made significant progress in understanding and...
TaskOpt: Dynamic Task Scheduling for Multi-Agent Systems | arxiv.org — Multi-agent systems have emerged as a promising approach to solving complex...