跳到正文
11月8日周日
11月7日周六
11月6日周五
  1. Rohan Paul53

    Nace AI 发布 Drex 开源决策模型技术报告,作者评价其用 27B 模型只做选择不做生成,专攻 Agent 中的大量小决策。Drex Max 在 Decision Bench v4 上与 GPT-6 Sol 统计持平,单次前向传播即输出各选项概率;成本 $0.06/1K 决策、延迟 183 ms,对比 Claude Opus 5.5 的 $10.17 和约 3.5 秒。

    引用Nace AI@NaceAI

    How are Decision Models Trained? Today, we are publishing the Drex Technical Report for the developer community! 🚀 Discover details on how the State-of-the-art Open-Source decision models are trained with RLAF. Build more at: console.nace.ai (try for free) P.S. Drex Max 1.1 and Drex for Finance are closed sourced and designed for enterprise use-cases

  2. IT之家·人工智能43

    webAI 开源 TwIL-8B 推理模型:聚焦数学与编程,长板表现优于 Gemma 4 31B

    webAI 开源了基于 IBM Granite-4.2-8B 的 TwIL-8B 推理模型,针对数学、代码和指令跟踪优化,采用非商业许可。该模型在 MATH-500、AIME 2025 两项数学基准上取得参比模型最佳成绩,强于尺寸数倍的 Gemma 4 31B,LiveCodeBench 同为同组最佳,IFBench 与 Gemma 4 12B 持平。

  3. 百度文心 ERNIE 官方博客75

    百度发布并开源 ERNIE-5.5-Flash-Base 预训练 Base 模型

    百度文心团队发布 ERNIE-5.5-Flash-Base,开放预训练终点 checkpoint,未经中训练与后训练,供社区自行开展领域适配、指令微调和强化学习。

    推荐理由:官方开放了未经后训练的预训练终点权重,并给出架构细节与同级 Base 模型评测,适合关注领域适配和继续训练的开发者参考。

  4. IT之家·人工智能52

    英伟达解读全栈协同设计破解 AI 内存瓶颈,每美元智能体吞吐量最高提升 30 倍

    英伟达 11 月 5 日发布技术解读,提出将 GPU、CPU、网络、存储和软件整体协同设计以缓解 HBM 供应趋紧带来的内存瓶颈。方案包括用 NVFP4 压缩权重和 KV 缓存、权重分散至最多 72 颗 GPU、Vera CPU 的 LPDDR 内存做就近缓冲、Dynamo 将不活跃上下文迁移至 CMX 存储。

  5. Rohan Paul58

    webAI 开源 TwIL-8B,一个基于 IBM Granite 4.2 8B 的 8.8B 参数模型,面向数学、编码和指令跟随。

    引用David Stout@Davidstout

    Modest intelligence is arriving faster than most people realize. Today, we're open-sourcing TwIL-8B. Just 8.8 billion parameters, yet approaching frontier-model scores on several difficult benchmarks: • 93% of frontier performance on coding • 92% on instruction following • 90% on contest mathematics All from a model small enough to run on your own hardware. But the benchmarks aren't the most important part of this announcement. For years, the AI industry has focused on building the most capable model, regardless of the cost required to run it. I believe that equation is changing. The next era of AI will be defined by the cost of completing work. Not simply the intelligence of the model, but how efficiently that intelligence can accomplish the task. If a model with 8 billion parameters can reliably complete the same work as one with hundreds of billions, why pay for the larger model? As smaller models become more capable, intelligence becomes something we can distribute, specialize, and run across the hardware we already own. That's the future we're building at @thewebAI TwIL-8B is available today. Open weights. And this is just the first of several announcements we're making today. HF Link: huggingface.co/webAI-Official…

11月5日周四
  1. Goodfire Research64

    Goodfire 提出 mechanistic shaping 框架 用内部表征抑制模型走捷径

    Goodfire Research 发布论文提出 mechanistic shaping 框架,通过 reliance shaping 和 expression shaping 在训练中惩罚模型对特定内部概念的依赖或表达,使模型以正确方式解题而非仅答对。

    推荐理由:论文提出用模型内部表征直接指定解题方式的训练框架,并系统评估了这么做之后监控探针如何退化与恢复。

  2. Rohan Paul51

    Stanford 研究者在 220 个真实专业任务上用同一模型 Qwen3.6-27B 测试 4 种单颗 AI 芯片,衡量每单位能量产生的有用工作。结果显示 NVIDIA B200 相比 AMD MI350X 每单位能量带来超过 4 倍的美元价值和超过 9 倍的专家时间节省,即模型不变、仅换芯片也会显著改变同样电力预算下的经济回报。作者引用的论文还构建了 Joules-to-GDP 指标,测试 5 个开源模型,其中 Gemma 4 31B 用 23% 的能量达到 GLM 5.2 每任务净价值的 56%,Qwen3.6-27B 从 1 卡 H100 扩到 8 卡仅快 22% 却耗能 3.2 倍。

    引用Rohan Paul@rohanpaul_ai

    New Stanford paper finds that a smaller AI model can return far more economic value per unit of energy than the most capable one. Now that the world is limited by energy/power, how much energy each AI task burns should indeed by studied. Here, the researchers built Joules-to-GDP, which measures the dollar value of an AI's work, minus its running cost, per joule of energy used. They tested 5 open models on 220 professional tasks from GDPval, covering 44 occupations. > GLM 5.2 produced the most net value per task, but Gemma 4 31B got 56% of it with 23% of the energy. > Moving Qwen3.6-27B from 1 to 8 H100s cut task time by just 22% while using 3.2x the energy. > GPT-OSS 120B used 68% less energy than GLM 5.2 despite nearly the same price and speed. > Running the same model on 8 H100 chips instead of 1 made tasks only 22% faster but used 3.24x the energy, for nearly the same dollar value of work x.com/jonsaadfalcon/status/21…

  3. Tomasz Tunguz 博客65

    Tomasz Tunguz 分析 2027 年 AI 算力天花板:80GW、GPU 涨价与效率主题

    Tomasz Tunguz 分析认为 2027 年美国 AI 算力上限约 80GW,即一年内新建相当于历史上全部算力规模的容量。短期 GPU 租金六个月内从每小时 $3 涨到 $12-13,Anthropic 混合 GPU 成本两个季度上升 82%,已建成机架因缺电无法启用;他推算理论收入上限接近 $500b,并提出效率将成为全栈主导主题。

11月4日周三
11月3日周二
  1. LlamaIndex:产品、工程与评测44

    OCRia-1 背后的推理栈:单张 B200 上 token 生成提速近 10 倍

    LlamaIndex 披露 OCRia-1 文档解析模型的推理优化方案,在单张 B200 GPU 上把 token 生成速度从约 285 tokens/sec 提升到约 2800 tokens/sec,接近 10 倍。方案结合了混合精度量化(BF16、FP8、NVFP4,视觉编码器不量化)与 DFlash2 块扩散投机解码,后者接受率超过 12 个 token。

  2. Rohan Paul63

    MIT CSAIL 研究者演示了 GPT-6 Astra 无需训练或梯度、直接读写神经网络权重的方法,称为 Weight-Space Programming。

    引用Rikiya Takehi@rikiyatakehi

    We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weight…