📡 外网新声
译文为机器翻译，供速览；点开原文为准。外链观点不代表本站立场。
- Hacker News刚收ENJev-rs: a Rust crate to turn a LLM to serve jevllm
- Hacker News刚收ENShow HN: Red-team your AI agentagent
- Hacker News刚收ENJev-serve: Run latest qwen3.8, other frozen LLM models, mlx-supportedllm
- Hacker News刚收译元测试模拟呼叫中心的人工智能座席呼叫agentMeta Tests Muse AI Agent Calls That Are Made by Humans in a Call Center
- Hacker News刚收ENUniversal Agent HP – 10-layer autonomous AI agent OS (free or local)agent
- Hacker NewsENShow HN: Life Forge – Open-source flight simulator for autonomous AI agentsagent
- Hacker NewsENTrained KV cache bank turns any LLM into Jev like Modelllm
- Hacker News译Meta的新Muse AI 智能体阅读我的私人消息。我从来没有要求过agentMeta's New Muse AI Agent Read My Private Messages. I Never Asked It To
- Hacker NewsENShow HN: Gwae – infinite-scroll terminal multiplexer for AI agent fleetsagent
- Hacker NewsENTradeoff considerations while running LLM models locallyllm
- Hacker NewsENNobodyWho is an inference engine that lets you run LLMs locally and efficientlyllm
- Hacker NewsENRecursive self-improvement of AI research agentsagent
- Hacker NewsENWhy Tool AIs Want to Be Agent AIsagent
- GitHub 新星精译anomalyco/opencode：开源编码智能体agentopen sourceanomalyco/opencode — The open source coding agent.
- GitHub 新星精译NousResearch/hermes-agent：与你一起成长的智能体agentNousResearch/hermes-agent — The agent that grows with you
- GitHub 新星精译ggml-org/llama.cpp：C/C++ 大模型推理引擎llmggml-org/llama.cpp — LLM inference in C/C++
- GitHub 新星精译earendil-works/pi：智能体工具箱（统一 LLM 接口·循环·终端界面·编码 CLI）agentllmearendil-works/pi — AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI
- GitHub 新星译DAO-AILab/flash-attention -快速且内存高效的精确注意memorydaoDao-AILab/flash-attention — Fast and memory-efficient exact attention
- GitHub 新星精译Shubhamsaboo/awesome-llm-apps：100+ 智能体 / 技能 / RAG 应用合集（开源）agentllmShubhamsaboo/awesome-llm-apps — 100+ AI Agents, Agent Skills and RAG Apps - Free and Open Source
- GitHub 新星译affaan-m/ECC —代理利用性能优化系统。技能、本能、记忆、安全、agentmemoryaffaan-m/ECC — The agent harness performance optimization system. Skills, instincts, memory, sec
- Hacker News精译ScopeTrail：为「多跳智能体委派」生成可审计回执agentagenticShow HN: ScopeTrail – audit receipts for multi-hop agent delegation
- GitHub 新星译obra/superpowers —有效的代理技能框架和软件开发方法。agentagenticobra/superpowers — An agentic skills framework & software development methodology that works.
- arXiv cs.AI译SpeakerMem-R1：以扬声器为中心的多方对话双轨内存memorySpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue
- arXiv cs.AI译CliffCompaction：长距离编码智能体的高性价比压缩agentCliffCompaction: Cost-Efficient Compaction for Long-Horizon Coding Agents
- arXiv cs.AI译SWE-Serve：对生产推理服务的代理工程进行基准测试agentagenticSWE-Serve: Benchmarking Agentic Engineering For Production Inference Serving
- arXiv cs.AI译A2M：MCP生态系统中的跟踪优化Agent劫持agentprotocolA2M: Trace-Optimized Agent Hijacking in the MCP Ecosystem
- arXiv cs.AI译发展线索，而不是背景：从无战略的支架到可重复使用的专业代理agentllmGrow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents
- arXiv cs.LGENEquivSVA: A Formally Verified Dataset of Behavioral Assertions Across Equivalent RTL Implementationslarge languagebenchmark
- arXiv cs.AI译基于LLM的代码漏洞修复中的指标失败：实证研究和变更感知屏幕llmlarge languageMetrics Failure in LLM-Based Code Vulnerability Repair: An Empirical Study and a Change-Aware Sc
- arXiv cs.LGENDiffusion-Induced Spatial Attention Overlapping Community Detectionneuraldiffusion
- arXiv cs.AI译人工智能是否节省了产品设计的时间？人工智能提示设计工作流程的随机对照实验llmlarge languageDoes AI Save Time on Product Design? A Randomized Controlled Experiment of AI Prompt-to-Design W
- arXiv cs.AI译《The Sirens 'Song: When Proximal Background Context Oversadows Distant Evidence》llmThe Sirens' Song: When Proximal Background Context Overshadows Distant Evidence
- arXiv cs.AIENTraceVIC: Causal Reasoning over Code Evolution for Identifying Vulnerability-Inducing Commitsreasoning
- arXiv cs.AIENTrain Where the Quantized Model Goes: On-Policy Distillation for Low-Bit Reasoningreasoning
- arXiv cs.LGENOptimal Sequential Annotations for Off-Policy Evaluationevaluationreinforcement
- arXiv cs.AIENBeyond Repeated Sampling: Learning Search Policies for LLM Reasoningllmlarge language
- arXiv cs.AIENMeasuring the Serving Stack Instead of the Model: Hidden Confounds in Local Tool-Use Evaluationagentevaluation
- arXiv cs.AIENFrom Alignment to Access Control: A Framework for GenAI Policy Enforcementagentlarge language
- arXiv cs.AIENA Spectral Theory of Grokking: Weight Decay induces Feature Learningneural
- arXiv cs.LGENMAGIC: Mixed-Granularity Agent Graphs via Incremental Construction with Dense-Reward Reinforcement Learningagentmulti-agent
- arXiv cs.LGENDiscovery-Driven Integration of Disjoint Tables via Textdataset
- arXiv cs.AIENThe Delegation Blind Spot: Auditing Product Decisions from Agent Choicesagent
- arXiv cs.AIENCapable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Modelsreasoning
- arXiv cs.LGENLabel-Efficient Learning for Ground-Based Sky-Image Classification: A Benchmark of Transfer Learning, Active Learning, and Pseudo-Labeling on GCDbenchmarkenergy
- arXiv cs.LGENOn Basis Function Selection for Sparse Gaussian Process Regressioncompute
- arXiv cs.AIENGreedy Decoding Is Not Precision-Invariant: Cross-Precision Output Divergence in LLM Inferencellmlarge language
- arXiv cs.LGENMMAP: Multimodal Missing-Aware Pretraining for Longitudinal Alzheimer's Predictionmultimodal
- GitHub 新星精译rasbt/LLMs-from-scratch：一步步用 PyTorch 从零实现类 ChatGPT 模型llmrasbt/LLMs-from-scratch — Implement a ChatGPT-like LLM in PyTorch from scratch, step by step
- OpenAI 官方译使用GPT ‑ 6 Astra将研究时间和成本并行削减一半agentParallel cut research time and cost in half with GPT‑6 Astra
- OpenAI 官方译有效第三方评估的优先事项和原则safetyPriorities and principles for effective third party assessments
- Hacker News译Show HN: Maki，开源多代理LLM框架（本地或托管）agentmulti-agentShow HN: Maki, an open-source multi-agent LLM framework (local or hosted)
- OpenAI 官方译为下一阶段的人工智能构建标准evaluationsafetyBuilding standards for the next phase of AI
- OpenAI 官方译V7如何为智能体提供机构记忆agentmemoryHow V7 gives AI agents institutional memory
- Hacker News译用于重现错误日志和打开PR的多代理工作流程agentmulti-agentMulti-agent workflows to reproduce error logs and open PRs
- Hacker News译HN，您好！memoryHi HN
- Hacker News译期待什么（多Agent系统的世界）agentmulti-agentWhat to Expect When You're Expecting (A World of Multi-Agent Systems)
- OpenAI 官方译澳大利亚青少年安全蓝图简介safetyIntroducing the Australian Youth Safety Blueprint
- GitHub 新星译mattpocock/skills —真正工程师的技能。直接来自我的.agents目录。agentmattpocock/skills — Skills for Real Engineers. Straight from my .agents directory.
- GitHub 新星译TauricResearch/TradingAgents — TradingAgents：多代理LLM金融交易框架agentmulti-agentTauricResearch/TradingAgents — TradingAgents: Multi-Agents LLM Financial Trading Framework
- OpenAI 官方译我们的模型错位报告框架alignmentOur framework for reporting model misalignment