已验证来源
每个故事都链接到最有力的可用证据:可用的原始来源,否则明确归因的报告。
简单的英语
发生了什么、为什么重要以及值得关注的内容——无需行话。
无填料
当信号很弱时,我们不会发布任何内容,而是填充提要。
更多故事
9 故事创新
FCPRAG 论文报告了参数 RAG 中检索到的证据更稳定的融合
EMNLP 2026 接受的一篇论文提出了 FCPRAG,这是一种控制器,可以在参数检索增强生成中选择性地组合特定于通道的 LoRA 适配器。作者报告称,在四个问答数据集中,F1 分数高于标准和参数 RAG 基线,增益高达 7.55%。arxiv.org创新
研究发现更多 GPU 资源并不能可靠地转化为更大的 NLP 研究影响
对主要 NLP 会议的 13,921 篇论文进行分析发现,报告最多 GPU 能力的论文捕获了大部分报告资源,但只有少数引用和奖项。该研究发现计算和学术影响之间存在统计关联,但独立的解释力很小。arxiv.org工业
筛选报告 Lunar 创始人为 AI 原生审计初创公司 Repodo 筹集 820 万欧元
Sifted 报道称,Lunar 前高管已筹集 820 万欧元种子前资金,用于推出 Repodo,这是一家人工智能原生法定审计初创公司,最初专注于中小企业。据 Sifted 数据显示,该轮融资是丹麦迄今为止最大的种子前融资。sifted.eu产品展示
chemeurope.com 报告 Skala AI 模型现已在 CP2K 中可用
chemeurope.com 报告称,Microsoft Research 的 Skala AI 模型已集成到开源 CP2K 模拟生态系统中,使研究人员能够使用基于 AI 的交换相关函数进行量子力学模拟。该媒体表示,初步测试显示特定测试的准确性有所提高……chemeurope.com工业
India Today reports Nvidia’s $6 billion Poolside deal expands its open-model ambitions
India Today reports that Nvidia’s agreement with AI startup Poolside would give Nvidia access to Poolside’s technology and bring more than 100 employees into its Nemotron open-weight model effort. The report also says Nvidia will invest $1 billion in Poolside at a $12 billion pre-money valuation, while rising…indiatoday.in创新
PUMA benchmark tests whether multimodal AI understands Polish culture
An arXiv preprint introduces PUMA, a 900-task benchmark for evaluating multimodal AI in Polish cultural and linguistic contexts, reporting strong visual question-answering performance but substantial weaknesses in complex audio and document understanding.arxiv.org创新
HiDiffTIR proposes difficulty-aware training for tool-using AI agents
A preprint introduces HiDiffTIR, a reinforcement-learning method that gives different weights to tool-use trajectories and reasoning steps based on their difficulty.arxiv.org工业
AI News reports XPENG robotics unit raises over $900 million at $6.3 billion valuation
AI News reports that XPENG’s physical AI unit has secured more than $900 million in funding at a $6.3 billion valuation to develop and commercialize its IRON humanoid robot.artificialintelligence-news.com创新
LURE paper proposes pursuit-evasion self-play to train LLM reasoning without task data
A new arXiv paper presents LURE, a zero-data self-play method in which one language model sets task difficulty while another solves verifiable reasoning challenges. The authors report stronger out-of-distribution zero-shot accuracy than trained baselines across nine held-out benchmarks, but the abstract does not…arxiv.org