OpenAI打造LLM超级黑客GPT-Red，与GPT-5.6对抗训练，让模型安全防御能力大幅提升。
OpenAI has built an LLM super-hacker called GPT-Red that it uses as a sparring partner to help its other models boost their defenses against cyberatta…
AI音乐生成明星Suno被曝利用YouTube数据训练，还隐瞒了安全泄露，行业伦理再敲警钟。
The hacker used an employee's credentials to access source code, which revealed how Suno scraped decades of audio.
专家训练时长如何影响LLM模型合并效果？这篇ICML 2026 workshop论文揭示关键发现。
arXiv:2607.11997v1 Announce Type: new Abstract: Multi-task model merging combines separately trained expert models into a single model that handles al…
出版商和作家组团起诉谷歌，指控其未经授权用版权作品训练Gemini AI，类似案件曾引发天价赔偿，版权与AI的博弈再升级。
IT之家 7 月 15 日消息，一群出版商和作家已对谷歌提起集体诉讼，指控这家科技巨头未经授权使用他们的版权作品训练其人工智能平台 Gemini。 原告方包括 Hachette Livre、Cengage Group、Elsevier、作家 Scott Turow 以及 S.C.R.I.B.E . …
Google内部文件泄露：用版权书籍训练AI面临百亿美元罚款，大型出版商再掀诉讼风暴。
Hachette, Cengage, Elsevier, and other publishers allege that Google trained its AI on copyrighted works without the necessary permissions.
Google 推出新版图片搜索功能，Spotify 开启语音交互功能内测等。 查看全文
通过合成指令数据扩展预训练规模，突破传统监督训练数据瓶颈的新方法
arXiv:2601.22146v2 Announce Type: replace-cross Abstract: Due to limited supervised training data, large language models (LLMs) are typically pre-trai…
一文对比RLHF与DPO两种主流大模型训练方法的核心差异与适用场景
In this article, we will look at how that learning actually happens, starting with why instruction-following alone falls short, then walking through t…
无需训练，LLM直接为度假租赁长尾房源生成推荐候选，破解协同过滤信号稀疏难题。
arXiv:2607.09877v1 Announce Type: new Abstract: Vacation rental marketplaces face a structural imbalance on the supply side: a small fraction of prope…
提出ARMOR方法，用离策锚点样本稳定在策强化学习训练大语言模型，解决训练震荡难题
arXiv:2607.10481v1 Announce Type: cross Abstract: Reinforcement learning (RL) has significantly enhanced the reasoning capabilities of large language …
微软CEO警告企业AI使用风险：你的核心数据可能正被模型制造商学习利用。
In a surprising blog post on Monday, Microsoft CEO is warning enterprises of the dangers of using proprietary models like Anthropic's and OpenAI's.
巧用物理规则为LLM提供连续奖励信号，让强化学习后训练更可解释、更高效
arXiv:2607.10474v1 Announce Type: cross Abstract: Partial differential equations (PDEs) are foundational to modeling in science and engineering, but c…
无需训练的LLM推理新方法：深度熵引导采样，提升推理效率与质量。
arXiv:2607.09693v1 Announce Type: cross Abstract: Reinforcement learning (RL) has become the dominant paradigm for improving the reasoning capabilitie…
高德发布ABot-World Studio，用通用世界模型工坊赋能智能体训练与模拟
36氪获悉，近日，阿里巴巴集团旗下高德正式发布通用世界模型工坊ABot-World Studio，并同步开放测试。该工坊将交互式视频生成与3DGS场景生成统一在同一产品中——用户只需输入一段文字或一张图片，即可生成一个可实时交互、任意分享的AI世界，输出结果可保存为视频与3DGS文件。
从训练营毕业生的视角，揭秘开源AI API的成本与体验，告诉你为何自建模型可能比API更贵。
Here's the thing: bootcamp Grad Explores Open-Source AI APIs: What I Learned I graduated from a coding bootcamp about six months ago, and honestly, I …
新论文提出UMoE方法，在领域特定训练中激活每个专家，突破传统MoE路径选择限制，提升模型适应性与效率。
arXiv:2607.11444v1 Announce Type: new Abstract: Mixture-of-Experts (MoE) models scale capacity without proportional compute cost and have become a key…
三星健康新增条款，用户拒绝AI训练授权将无法备份数据，健康数据与功能捆绑引发隐私争议。
IT之家 7 月 14 日消息，科技媒体 Neowin 昨日（7 月 13 日）发布博文，报道称三星健康（Samsung Health）新增选项， 用户若不同意其健康数据用于训练 AI，可能无法备份数据且被威胁删除数据。 在三星健康应用设置中，三星新增“同意将健康数据用于人工智能训练和建模”的选项，…
从零训练113M参数地震大模型，开源全流程管线，自建数据集+分布式训练，值得参考复现。
Article URL: https://github.com/jiazhe868/nanogpt-seis Comments URL: https://news.ycombinator.com/item?id=48885236 Points: 9 # Comments: 2
无需隐变量，三值量化LLM训练新方法，开源项目BitBop让你以极低成本体验高效训练。
Article URL: https://github.com/ValerioDolci/bitbop Comments URL: https://news.ycombinator.com/item?id=48892013 Points: 1 # Comments: 1
巧妙利用预训练模型权重谱的结构化模式作为初始化信号，提升GPT-2风格语言模型预训练效率
arXiv:2607.09204v1 Announce Type: cross Abstract: Pretrained language models often exhibit structured weight spectra, suggesting that training may rep…