Skip to content

AI论文速递 2026年08月18日（HuggingFace Daily Papers）

数据来源：https://huggingface.co/papers 采集时间：2026-08-18

📌 重点关注

  1. MobileMem: Learning from a Year of Mobile Experiences | arXiv — 【重点关注】 The next generation of AI agents is increasingly moving beyond systems that a... 💡 移动端Agent长期记忆，端侧AI与移动经验结合的落地范本
  2. Multimodal Model Diffing for Feature Discovery and Control | arXiv — 【重点关注】 Multimodal Large Language Models (MLLMs) exhibit strong visual understanding,... 💡 模型差分发现/控制MLLM特征，可解释性新思路，可借鉴做诊断
  3. Second Thought: Reasoning in Parallel as LLM Agents Act and Observe | arXiv — 【重点关注】 LLM agents in the ReAct paradigm alternate between reasoning, acting, and obs... 💡 ReAct并行推理提速Agent，对Agent架构优化有直接参考价值

📋 其他值得关注

  1. VibeWorlding: Can Multimodal Agents Construct 3D Open Worlds End-to-End? | arXiv — Constructing an interactive 3D open world from a user query is important. How...
  2. Agentic Transaction: Towards ACID-Compliant Agent Systems | arXiv — Large language model (LLM) agents are evolving from conversational assistants...
  3. R^3-Bench: LLMs Struggle with Resource-Rational Reasoning under Shared Budgets | arXiv — In cognitive science, resource rationality asks how an agent should allocate ...
  4. Hybrid-Policy Self-Editing for Composable Unstructured Knowledge Editing | arXiv — Large language models (LLMs) achieve remarkable performance across natural la...
  5. GenRouter: Unified Workflow Routing for Agentic Image Generation | arXiv — The rapid evolution of text-to-image (T2I) generation models has effectively ...
  6. Latent On-Policy Self-Distillation | arXiv — Enabling agents to learn from experience and internalize it into their policy...
  7. Amplified Does Not Mean Predictive: Reasoning Behaviors in Thinking Models | arXiv — Which reasoning behaviors are associated with correct answers in reasoning mo...