发布日期：2026-05-17
收录条目：20
1. Nous Research Proposes Lighthouse Attention: A Training-Only Selection-Based Hierarchical Attention That Delivers 1.4–1.7× Pretraining Speedup at Long Context
- 来源：MarkTechPost
- 发布时间：2026-05-16 22:23 UTC
- 链接：https://www.marktechpost.com/2026/05/16/nous-research-proposes-lighthouse-attention-a-training-only-selection-based-hierarchical-attention-that-delivers-1-4-1-7x-pretraining-speedup-at-long-context/
摘要：Nous Research has published Lighthouse Attention, a selection-based hierarchical attention mechanism that wraps around standard scaled dot-product attention during pretraining and is removed afterward. Unlike prior metho
2. Meet LiteLLM Agent Platform: A Kubernetes-Based, Self-Hosted Infrastructure Layer for Isolated Agent Sandboxes and Persistent Session Management in Production
- 来源：MarkTechPost
- 发布时间：2026-05-16 17:59 UTC
- 链接：https://www.marktechpost.com/2026/05/16/meet-litellm-agent-platform-a-kubernetes-based-self-hosted-infrastructure-layer-for-isolated-agent-sandboxes-and-persistent-session-management-in-production/
摘要：Running AI agents in a local script is straightforward. Running them reliably in production across teams, across restarts, with isolated environments per context is a different problem entirely. BerriAI, the company behi
3. Sony tries to explain that its AI Camera Assistant doesn’t suck
- 来源：The Verge AI
- 发布时间：2026-05-16 15:37 UTC
- 链接：https://www.theverge.com/tech/932133/sony-xperia-1-xiii-ai-camera-assistant
摘要：After Sony drew some unwanted attention for a post demonstrating its AI Camera Assistant on the Xperia 1 XIII, it's trying to clarify how the feature works. The company says it doesn't edit photos, but makes suggestions
4. NVIDIA Introduces SANA-WM: A 2.6B-Parameter Open-Source World Model That Generates Minute-Scale 720p Video on a Single GPU
- 来源：MarkTechPost
- 发布时间：2026-05-16 07:52 UTC
- 链接：https://www.marktechpost.com/2026/05/16/nvidia-introduces-sana-wm-a-2-6b-parameter-open-source-world-model-that-generates-minute-scale-720p-video-on-a-single-gpu/
摘要：Researchers from NVIDIA introduce SANA-WM, an open-source camera-controlled world model that generates 60-second, 720p videos with precise 6-DoF camera control — trained on 64 H100 GPUs and deployable on a single RTX 509
5. How to Build Repository-Level Code Intelligence with Repowise Using Graph Analysis, Dead-Code Detection, Decisions, and AI Context
- 来源：MarkTechPost
- 发布时间：2026-05-16 06:45 UTC
- 链接：https://www.marktechpost.com/2026/05/15/how-to-build-repository-level-code-intelligence-with-repowise-using-graph-analysis-dead-code-detection-decisions-and-ai-context/
摘要：In this tutorial, we explore how to use Repowise to build repository-level intelligence for the itsdangerous Python project in a practical and reproducible way. We start with an already cloned repository, configure Repow
6. GraphBit: A Graph-based Agentic Framework for Non-Linear Agent Orchestration
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.13848
摘要：arXiv:2605.13848v1 Announce Type: new Abstract: Agentic LLM frameworks that rely on prompted orchestration, where the model itself determines workflow transitions, often suffer from hallucinated routing, infinite loops,
7. Mixed Integer Goal Programming for Personalized Meal Optimization with User-Defined Serving Granularity
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.13849
摘要：arXiv:2605.13849v1 Announce Type: new Abstract: Determining what to eat to satisfy nutritional requirements is one of the oldest optimization problems in operations research, yet existing formulations have two persistent
8. A Two-Dimensional Framework for AI Agent Design Patterns: Cognitive Function and Execution Topology
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.13850
摘要：arXiv:2605.13850v1 Announce Type: new Abstract: Existing frameworks for LLM-based agent architectures describe systems from a single perspective: industry guides (Anthropic, Google, LangChain) focus on execution topology
9. Invisible Orchestrators Suppress Protective Behavior and Dissociate Power-Holders: Safety Risks in Multi-Agent LLM Systems
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.13851
摘要：arXiv:2605.13851v1 Announce Type: new Abstract: Multi-agent orchestration -- in which a hidden coordinator manages specialized worker agents -- is becoming the default architecture for enterprise AI deployment, yet the s
10. PREPING: Building Agent Memory without Tasks
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.13880
摘要：arXiv:2605.13880v1 Announce Type: new Abstract: Agent memory is typically constructed either offline from curated demonstrations or online from post-deployment interactions. However, regardless of how it is built, an age
11. PolitNuggets: Benchmarking Agentic Discovery of Long-Tail Political Facts
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14002
摘要：arXiv:2605.14002v1 Announce Type: new Abstract: Large Reasoning Models (LRMs) embedded in agentic frameworks have transformed information retrieval from static, long context question answering into open-ended exploration
12. Conditional Attribute Estimation with Autoregressive Sequence Models
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14004
摘要：arXiv:2605.14004v1 Announce Type: new Abstract: Generative models are often trained with a next-token prediction objective, yet many downstream applications require the ability to estimate or control sequence-level prope
13. Sheaf-Theoretic Transport and Obstruction for Detecting Scientific Theory Shift in AI Agents
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14033
摘要：arXiv:2605.14033v1 Announce Type: new Abstract: Scientific theory shift in AI agents requires more than fitting equations to data. An artificial scientific agent must detect whether an existing representational framework
14. From Descriptive to Prescriptive: Uncover the Social Value Alignment of LLM-based Agents
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14034
摘要：arXiv:2605.14034v1 Announce Type: new Abstract: Wide applications of LLM-based agents require strong alignment with human social values. However, current works still exhibit deficiencies in self-cognition and dilemma dec
15. Enhanced and Efficient Reasoning in Large Learning Models
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14036
摘要：arXiv:2605.14036v1 Announce Type: new Abstract: In current Large Language Models we can trust the production of smoothly flowing prose on the basis of the principles of machine learning. However, there is no comparably p
16. Model-Adaptive Tool Necessity Reveals the Knowing-Doing Gap in LLM Tool Use
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14038
摘要：arXiv:2605.14038v1 Announce Type: new Abstract: Large language models (LLMs) increasingly act as autonomous agents that must decide when to answer directly vs. when to invoke external tools. Prior work studying adaptive
17. Network-Aware Bilinear Tokenization for Brain Functional Connectivity Representation Learning
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14048
摘要：arXiv:2605.14048v1 Announce Type: new Abstract: Masked autoencoders (MAEs) have recently shown promise for self-supervised representation learning of resting-state brain functional connectivity (FC). However, a fundament
18. Bridging Legal Interpretation and Formal Logic: Faithfulness, Assumption, and the Future of AI Legal Reasoning
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14049
摘要：arXiv:2605.14049v1 Announce Type: new Abstract: The growing adoption of large language models in legal practice brings both significant promise and serious risk. Legal professionals stand to benefit from AI that can reas
19. SPIN: Structural LLM Planning via Iterative Navigation for Industrial Tasks
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14051
摘要：arXiv:2605.14051v1 Announce Type: new Abstract: Industrial LLM agent systems often separate planning from execution, yet LLM planners frequently produce structurally invalid or unnecessarily long workflows, leading to br
20. Bad Seeing or Bad Thinking? Rewarding Perception for Vision-Language Reasoning
- 来源：arXiv cs.AI
- 发布时间：2026-05-16 04:00 UTC
- 链接：https://arxiv.org/abs/2605.14054
摘要：arXiv:2605.14054v1 Announce Type: new Abstract: Achieving robust perception-reasoning synergy is a central goal for advanced Vision-Language Models (VLMs). Recent advancements have pursued this goal via architectural des