AI论文速递 2026年08月15日（HuggingFace Daily Papers）

数据来源：https://huggingface.co/papers 采集时间：2026-08-15

📌 重点关注

1. Spatial Memory Agent: Experience-Grounded Procedure Memory for Spatial Intelligence | arXiv — 【重点关注】 Spatial intelligence is becoming a foundation for embodied agents, robotic pl... 💡 具身智能空间记忆落地范式，对端侧Agent记忆设计有借鉴
2. Intern-S2-Preview: Scientific Agentic Foundation Model | arXiv — 【重点关注】 Scientific discovery increasingly requires AI systems that can reason over sc... 💡 科学Agent基座，展现多步推理与工具编排的工业级范式
3. LycheeMemory V2: Efficient Long-Term Memory for LLM Agents via Semantic Segment-Level Consolidation | arXiv — 【重点关注】 Long-horizon LLM agents must preserve information from past interactions to s... 💡 语义分段级记忆巩固，直接可借鉴到个人知识库长程记忆

📋 其他值得关注

1. AutoDesign: Meta-Harness Optimization for Long-Horizon Agentic Design | arXiv — Transforming multimodal sources into condensed and structured media outputs c...
2. The Illusion of Visual Tool-Use: A Causal Audit of Thinking with Images | arXiv — The "thinking-with-images" paradigm equips multimodal LLMs with active visual...
3. DarwinX: Evolving Agent Harnesses Through Natural Selection | arXiv — An LLM agent's capability depends not only on model weights but on its harnes...
4. OmniScientist: An Omni-Modal Omni-Discipline AI Scientist | arXiv — Recent advances in foundation models have enabled AI scientists to automate i...
5. Hybrid-Policy Self-Editing for Composable Unstructured Knowledge Editing | arXiv — Large language models (LLMs) achieve remarkable performance across natural la...
6. Parameter Exploration for RLVR via Variational Learning | arXiv — Exploration has been a focus of reinforcement learning research for a long ti...
7. Self-Evolving Embodied Agents via Skill-Harness Evolution | arXiv — Embodied agents are increasingly built as systems around foundation models, w...