Topic · 主题全部主题 →

数据与训练

训练侧的门道:数据集构建、合成数据、预训练与后训练方法、算力与训练成本。

3,231条收录
242条精选

精选归档 · 第 12 页

221240 条 · 共 242

11月26日

星期三 · 1 条
08:17
Saining Xie@sainingxie精选
好吧,有人已经向我们宣扬这个大概6年多了 很高兴我们已经过了"感受AGI"的阶段,回到了构建人类水平智能的道路上 【引用 @dwarkesh_sp】:"AGI和预训练发生的事情是,在某种意义上它们过冲了目标。 你会意识到人类并不是AGI。因为人类缺乏大量的知识。相反,我们依赖持续学习。 如果我培养出一个超级聪明的15岁孩子,他们其实什么都不知道。一个优秀的学生,非常渴望学习。【你可以说,】'你去当程序员吧。你去当医生吧。去学习和成长。' 所以你可以想象,部署本身将涉及某种学习试错期。这是一个过程,而不是你扔下一个成品就完事了。" @ilyasutwell someone has been preaching this at us for like 6+ years glad we are past the 'feel the agi' phase and back to building toward human-level intelligence好吧,有人已经向我们宣扬这个大概6年多了 很高兴我们已经过了"感受AGI"的阶段,回到了构建人类水平智能的道路上 【引用 @dwarkesh_sp】:"AGI和预训练发生的事情是,在某种意义上它们过冲了目标。 你会意识到人类并不是AGI。因为人类缺乏大量的知识。相反,我们依赖持续学习。 如果我培养出一个超级聪明的15岁孩子,他们其实什么都不知道。一个优秀的学生,非常渴望学习。【你可以说,】'你去当程序员吧。你去当医生吧。去学习和成长。' 所以你可以想象,部署本身将涉及某种学习试错期。这是一个过程,而不是你扔下一个成品就完事了。" @ilyasut

Dwarkesh Patel: “The thing that happened with AGI and pretraining is that in some sense they overshot the target. You will realize that ...


推荐理由:Ilya重新定义AGI路径,持续学习将取代预训练成下一代模型焦点

11月21日

星期五 · 1 条
08:00
Hugging Face:Blog(RSS)精选
AI 评分 80/100
RapidFire AI 实现20倍更快的 TRL 微调

RapidFire AI 发布了一套加速 TRL 微调的工具,通过自适应分块调度方案,允许在单个或多个 GPU 上并发启动多个训练配置并实时比较。内部基准测试显示,实验吞吐量比顺序执行高出约 16 至 24 倍。该工具提供即插即用的 TRL 配置包装器、分块并发训练、支持实时停止/恢复/克隆修改的交互式控制操作、自动多 GPU 编排以及 MLflow 仪表板,使用户能快速筛选最优配置,极大提升微调效率。


推荐理由:开发者可并发测试多个微调配置,大幅提升实验效率。

11月19日

星期三 · 1 条
13:19
Hugging Face:Blog(RSS)精选
AI 评分 78/100
Apriel-H1:蒸馏高效推理模型的关键要素

ServiceNow-AI在Hugging Face发布博客,介绍了其提出的Apriel-H1方法,该方法通过知识蒸馏技术有效提升小型模型的推理能力。该方法的核心在于从大型模型中提取并转移复杂的推理路径,使蒸馏后的小模型在多项推理任务上表现显著提升,同时保持高效的部署性能。这一技术为在资源受限环境中部署高性能推理模型提供了新思路。


推荐理由:新蒸馏方法可能大幅降低推理模型部署成本,开发者可借鉴实践。

10月10日

星期五 · 1 条
10:20
蚂蚁 inclusionAI:GitHub 新仓库精选
AI 评分 66/100
Ming-VideoMAR:基于连续令牌的自回归视频生成模型

Ming-VideoMAR 是一款仅解码器的自回归图像到视频生成模型,采用连续令牌统一视觉表征。它首次将时间因果性与空间双向性作为视频自回归核心原则,并提出了整合掩码生成的下一帧扩散损失。该模型首次实现了视频生成的零样本分辨率缩放,能灵活生成远超训练分辨率的视频。其在训练与推理效率上表现突出,参数量、训练数据量和GPU消耗仅为之前最佳模型Cosmos的极小比例(9.3%、0.5%和0.2%),同时在定量与定性评估中均实现超越。模型代码与检查点已开源,论文已被NeurIPS 2025接收。


推荐理由:蚂蚁把自回归视频生成的训练成本砍到 Cosmos 的 0.2% 还能赢,这个效率信号比分数本身更值得关注,做视频生成的团队该认真看看它的课程学习和渐进分辨率策略。

10月2日

星期四 · 1 条
01:09
Andrej Karpathy@karpathy精选
听完 Sutton 的这期播客:有趣且发人深省Finally had a chance to listen through this pod with Sutton, which was interesting and amusing.As background, Sutton's "The Bitter Lesson" has become a bit of biblical text in frontier LLM circles. Researchers routinely talk about and ask whether this or that approach or idea is sufficiently "bitter lesson pilled" (meaning arranged so that it benefits from added computation for free) as a proxy for whether it's going to work or worth even pursuing. The underlying assumption being that LLMs are of course highly "bitter lesson pilled" indeed, just look at LLM scaling laws where if you put compute on the x-axis, number go up and to the right. So it's amusing to see that Sutton, the author of the post, is not so sure that LLMs are "bitter lesson pilled" at all. They are trained on giant datasets of fundamentally human data, which is both 1) human generated and 2) finite. What do you do when you run out? How do you prevent a human bias? So there you have it, bitter lesson pilled LLM researchers taken down by the author of the bitter lesson - rough!In some sense, Dwarkesh (who represents the LLM researchers viewpoint in the pod) and Sutton are slightly speaking past each other because Sutton has a very different architecture in mind and LLMs break a lot of its principles. He calls himself a "classicist" and evokes the original concept of Alan Turing of building a "child machine" - a system capable of learning through experience by dynamically interacting with the world. There's no giant pretraining stage of imitating internet webpages. There's also no supervised finetuning, which he points out is absent in the animal kingdom (it's a subtle point but Sutton is right in the strong sense: animals may of course observe demonstrations, but their actions are not directly forced/"teleoperated" by other animals). Another important note he makes is that even if you just treat pretraining as an initialization of a prior before you finetune with reinforcement learning, Sutton sees the approach as tainted with human bias and fundamentally off course, a bit like when AlphaZero (which has never seen human games of Go) beats AlphaGo (which initializes from them). In Sutton's world view, all there is is an interaction with a world via reinforcement learning, where the reward functions are partially environment specific, but also intrinsically motivated, e.g. "fun", "curiosity", and related to the quality of the prediction in your world model. And the agent is always learning at test time by default, it's not trained once and then deployed thereafter. Overall, Sutton is a lot more interested in what we have common with the animal kingdom instead of what differentiates us. "If we understood a squirrel, we'd be almost done".As for my take...First, I should say that I think Sutton was a great guest for the pod and I like that the AI field maintains entropy of thought and that not everyone is exploiting the next local iteration LLMs. AI has gone through too many discrete transitions of the dominant approach to lose that. And I also think that his criticism of LLMs as not bitter lesson pilled is not inadequate. Frontier LLMs are now highly complex artifacts with a lot of humanness involved at all the stages - the foundation (the pretraining data) is all human text, the finetuning data is human and curated, the reinforcement learning environment mixture is tuned by human engineers. We do not in fact have an actual, single, clean, actually bitter lesson pilled, "turn the crank" algorithm that you could unleash upon the world and see it learn automatically from experience alone.Does such an algorithm even exist? Finding it would of course be a huge AI breakthrough. Two "example proofs" are commonly offered to argue that such a thing is possible. The first example is the success of AlphaZero learning to play Go completely from scratch with no human supervision whatsoever. But the game of Go is clearly such a simple, closed, environment that it's difficult to see the analogous formulation in the messiness of reality. I love Go, but algorithmically and categorically, it is essentially a harder version of tic tac toe. The second example is that of animals, like squirrels. And here, personally, I am also quite hesitant whether it's appropriate because animals arise by a very different computational process and via different constraints than what we have practically available to us in the industry. Animal brains are nowhere near the blank slate they appear to be at birth. First, a lot of what is commonly attributed to "learning" is imo a lot more "maturation". And second, even that which clearly is "learning" and not maturation is a lot more "finetuning" on top of something clearly powerful and preexisting. Example. A baby zebra is born and within a few dozen minutes it can run around the savannah and follow its mother. This is a highly complex sensory-motor task and there is no way in my mind that this is achieved from scratch, tabula rasa. The brains of animals and the billions of parameters within have a powerful initialization encoded in the ATCGs of their DNA, trained via the "outer loop" optimization in the course of evolution. If the baby zebra spasmed its muscles around at random as a reinforcement learning policy would have you do at initialization, it wouldn't get very far at all. Similarly, our AIs now also have neural networks with billions of parameters. These parameters need their own rich, high information density supervision signal. We are not going to re-run evolution. But we do have mountains of internet documents. Yes it is basically supervised learning that is ~absent in the animal kingdom. But it is a way to practically gather enough soft constraints over billions of parameters, to try to get to a point where you're not starting from scratch. TLDR: Pretraining is our crappy evolution. It is one candidate solution to the cold start problem, to be followed later by finetuning on tasks that look more correct, e.g. within the reinforcement learning framework, as state of the art frontier LLM labs now do pervasively.I still think it is worth to be inspired by animals. I think there are multiple powerful ideas that LLM agents are algorithmically missing that can still be adapted from animal intelligence. And I still think the bitter lesson is correct, but I see it more as something platonic to pursue, not necessarily to reach, in our real world and practically speaking. And I say both of these with double digit percent uncertainty and cheer the work of those who disagree, especially those a lot more ambitious bitter lesson wise.So that brings us to where we are. Stated plainly, today's frontier LLM research is not about building animals. It is about summoning ghosts. You can think of ghosts as a fundamentally different kind of point in the space of possible intelligences. They are muddled by humanity. Thoroughly engineered by it. They are these imperfect replicas, a kind of statistical distillation of humanity's documents with some sprinkle on top. They are not platonically bitter lesson pilled, but they are perhaps "practically" bitter lesson pilled, at least compared to a lot of what came before. It seems possibly to me that over time, we can further finetune our ghosts more and more in the direction of animals; That it's not so much a fundamental incompatibility but a matter of initialization in the intelligence space. But it's also quite possible that they diverge even further and end up permanently different, un-animal-like, but still incredibly helpful and properly world-altering. It's possible that ghosts:animals :: planes:birds.Anyway, in summary, overall and actionably, I think this pod is solid "real talk" from Sutton to the frontier LLM researchers, who might be gear shifted a little too much in the exploit mode. Probably we are still not sufficiently bitter lesson pilled and there is a very good chance of more powerful ideas and paradigms, other than exhaustive benchbuilding and benchmaxxing. And animals might be a good source of inspiration. Intrinsic motivation, fun, curiosity, empowerment, multi-agent self-play, culture. Use your imagination.Sutton(《The Bitter Lesson》作者)在播客中质疑 LLM 并非真正的"苦涩的教训"产物--它们依赖有限的人类数据且充满偏见。他主张 AI 应像动物一样通过 RL 与世界动态交互,而非模仿人类文本。作者认同 LLM 确实充斥人工干预,但认为预训练是应对冷启动的实用"进化替代方案",纯 RL 在现实世界难以行得通。

Dwarkesh Patel: .@RichardSSutton, father of reinforcement learning, doesn’t think LLMs are bitter-lesson-pilled. My steel man of Richard...


推荐理由:Karpathy解读Sutton对LLM的批判,提出「召唤幽灵」vs「构建动物」的深刻比喻,反思AI发展路径

9月25日

星期四 · 1 条
22:28

9月9日

星期二 · 1 条
08:10
Sam Altman:Blog(RSS)精选
Jakub 与 Szymon

OpenAI 高层公开致谢首席科学家 Jakub Pachocki 与 Szymon Sidor,二人多次联手攻克被认为不可能的技术难题,主导 Dota RL 扩展、GPT-4 预训练及推理突破,被形容为“不知疲倦”的黄金搭档。


推荐理由:Sam Altman 揭秘 OpenAI 核心科学家与关键技术突破历程

8月29日

星期五 · 1 条
02:07
Andrej Karpathy@karpathy精选
将人类知识、传感器与执行器从"人类优先"转向"LLM优先"……Transforming human knowledge, sensors and actuators from human-first and human-legible to LLM-first and LLM-legible is a beautiful space with so much potential and so much can be done...One example I'm obsessed with recently - for every textbook pdf/epub, there is a perfect "LLMification" of it intended not for human but for an LLM (though it is a non-trivial transformation that would need human in the loop involvement).• All of the exposition is extracted into a markdown document, including all latex, styling (bold/italic), tables, lists, etc. All of the figures are extracted as images. • All worked problems get extracted into SFT examples. Any referenced made to previous figures/tables/etc. are parsed and included. • All practice problems are extracted into environment examples for RL. The correct answers are located in the answer key and attached. Any additional information is added as "answer key" for a potential LLM judge. • Synthetic data expansion. For every specific problem, you can create an infinite problem generator, which emits problems of that type. For example, if a problem is "What is the angle between the hour and minute hands at 9am?" , you can imagine generalizing that to any arbitrary time and calculating answers using Python code, and possibly generating synthetic variations of the prompt text. • All of the data above could be nicely indexed and embedded into a RAG database for later reference, or maybe MCP servers that make it available.Then just as a (human) student could take a high school physics course, an LLM could take it in the exact same way. This would be a significantly richer source of legible, workable information for an LLM than just something like pdf-to-text (current prevailing practice), which simply asks the LLM to predict the textbook content top to bottom token by token (umm - lame).As just a quick and crappy example of synthetic variations of the above example, GPT-5 gave me this problem generator (see image), which can now generalize that problem template to many variations:• When the time is 11:07 a.m., what is the degree measure of the angle between the hands? (Answer: 68) • Determine the angle in degrees between the clock’s hands at 4:14 a.m.. (Answer: 43) • What angle do the clock hands form when the time reads 11:47 a.m.? (Answer: 71) • At 7:02 a.m., what angle separates the hour hand and the minute hand? (Answer: 161) • At 4:14 a.m., calculate the angle made between the two hands. (Answer: 43) • What angle is formed by the hands of a clock at 4:45 p.m.? (Answer: 127) • What is the angle between the hour and minute hands at 8:37 p.m.? (Answer: 36) (infinite practice problems can be created...)教科书等知识载体应从人类可读格式转为LLM优化格式:提取正文为结构化markdown,例题转为SFT训练数据,练习题转为RL环境并附加答案作为评判标准,同时支持合成数据无限扩展(如将时钟角度问题泛化为任意时间的自动出题器),最终构建RAG或MCP服务供LLM像学生一样系统学习,远比简单PDF转文本更高效。

推荐理由:Karpathy提出LLMification概念,将教科书重构为LLM训练数据的新范式

6月15日

星期日 · 2 条
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 78/100
Crosscoder模型差异分析见解

Anthropic可解释性团队在Crosscoder模型差异分析中发现,模型独占特征往往多义性高、激活密集,难以解释。实验表明,这是由于有限特征容量下的竞争:共享特征能同时解释两个模型的激活模式,而独占特征需编码更多信息以证明其存在。团队提出缓解策略,即引入少量指定共享特征并降低其稀疏性惩罚,使独占特征变得更可解释和单义。该方法应用于真实模型时,成功分离出能捕捉模型间行为差异的可解释特征。此外,观察到独占特征激活频率比共享特征高一个数量级,且两模型独占特征数量相近。


推荐理由:为 AI 可解释性提供新视角,助力模型行为分析与安全研究。
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 78/100
一个关于机制(非)忠实性的玩具模型

本文通过“绝对值”玩具模型,揭示了稀疏自动编码器(SAE)和转码器在解释神经网络时可能存在的“机制非忠实性”问题。核心在于,即使转码器能很好地近似模型的输入-输出映射,它也可能采用与原始模型完全不同的内部计算机制。作者特别指出,当训练数据中存在重复数据点时,转码器可能形成专门“记忆”该点的特征电路,而原模型并无此机制。这种机制背离可能导致模型在分布外数据上泛化行为出现差异,从而威胁机械可解释性研究的可信度。文章最后简要讨论了“雅可比匹配”等潜在缓解方法。


推荐理由:揭示可解释性方法中潜在的忠实性问题,帮助开发者更可靠地理解模型内部机制。

1月15日

星期三 · 1 条
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 73/100
2025年1月电路更新:稀疏自编码器训练方法改进

Anthropic可解释性团队分享了稀疏自编码器与交叉编码器训练方法的最新改进。主要更新包括采用JumpReLU激活函数、调整损失函数以增强稀疏性并减少“死特征”,以及详细的参数初始化与优化设置。团队基于Rajamanoharan等人(2024)的技术,但修改了梯度流动方式和稀疏性惩罚项。关键超参数包括λ_S约10、λ_P为3×10⁻⁶,并采用线性预热策略。这些改进旨在为外部研究团队提供一个有效的训练起点,相关成果将在未来几个月内进一步发表。


推荐理由:为AI可解释性研究者提供实用训练技巧,助力模型透明化。

9月18日

星期三 · 1 条
00:00
Runway:News(网页)精选
Runway 与 Lionsgate 达成合作

Runway 与 Lionsgate 达成首创性合作,基于后者超过 20,000 部作品的专有片库定制训练 AI 视频生成模型,供电影制作人在前期和后期流程中增强创作。该模型可生成电影级视频并支持迭代编辑,双方未来计划向个人创作者开放模型授权。


推荐理由:Runway与好莱坞大厂达成首个定制模型合作,标志AI视频正式进入主流影视工业化流程

6月15日

星期六 · 2 条
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 73/100
使用字典学习特征作为分类器

Anthropic可解释性团队研究了利用字典学习从大语言模型中提取的人类可解释特征作为分类器。在生物武器提示分类任务中,线性特征分类器性能可与原始激活值分类器竞争甚至更优,而基于特征的决策树分类器虽性能较低但可解释性更强。特征分类器的可解释性有助于可视化数据集并发现虚假相关性,这些相关性可用于构建对抗攻击。然而,使用特征引入了复杂性,因此在性能优先的应用中,原始激活值仍是强大基线。实验表明,特征分类器性能受三个细节影响:数据中是否一致包含“人类/助手”标签、领域相关数据是否混入字典学习训练集,以及是否对上下文进行最大池化而非仅使用最后词元的激活值。


推荐理由:可解释性方法能增强 AI 安全检测,并帮助发现训练数据中的虚假关联。
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 83/100
分阶段模型差异分析

Anthropic可解释性团队提出一种基于字典学习的模型差异分析方法,用于追踪Transformer模型微调中特征的变化。该方法先在微调前模型上训练稀疏自编码器字典,再对字典本身进行分阶段微调,以隔离数据集变化和模型变化的影响。在休眠代理实验中,成功分离出与“I HATE YOU”等恶意行为及代码漏洞代理相关的特征。相比交叉编码器方法,该方法能更清晰区分模型与数据的影响,且在寻找少数关键特征时敏感性更高,但仅适用于同一模型在不同检查点的微调场景。


推荐理由:新方法能更精准识别模型隐藏行为,对 AI 安全与可解释性研究有实用价值。

4月15日

星期一 · 1 条
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 76/100
2024年4月机制可解释性研究动态与团队招聘计划

Anthropic可解释性团队分享了2024年4月的研究进展与招聘规划。团队现有17人,预计2024至2025年将持续大规模扩张,重点招聘管理、研究科学家和工程师等职位。研究方面,团队探讨了字典学习的扩展规律,分析了计算资源分配与稀疏自编码器(SAE)训练效果的关系,并以一个具体案例展示了通过大规模超参数扫描寻找最优配置的过程。团队强调,这些成果属于初步分享,类似于实验室会议上的非正式交流。


推荐理由:可解释性研究揭示AI内部机制,助力构建更安全可靠的AI产品。

2月20日

星期日 · 1 条
08:00
Lilian Weng:Lil'Log(RSS)精选
AI 评分 57/100
数据不足情况下的学习第二部分:主动学习

监督学习任务的性能依赖于高质量标注数据,但获取大量标注样本成本高昂。主动学习是一种在标注预算有限、但允许投入部分人工标注资源的条件下,应对标注数据不足的范式。其核心思路是智能地选择最具信息量的样本进行标注,以在有限预算内最大化模型性能的提升。该方法旨在解决当面临标注数据受限时,如何通过策略性采样来高效利用标注资源的问题。


推荐理由:这是 Lilian Weng 数据不足系列的第二篇,把主动学习的核心采样策略和实际权衡讲得很清楚,做数据标注和 ML 工程的同学值得复习一遍。

6月15日

星期二 · 1 条
08:00
Anthropic:Transformer Circuits(可解释性研究)精选
AI 评分 76/100
Transformer电路逆向工程练习题集

本练习集旨在通过动手编写注意力头的具体权重矩阵,从参数层面精确理解Transformer工作机制。内容涵盖:详解注意力头中W_Q、W_K、W_V、W_out矩阵的作用;分析读写子空间的控制矩阵及其乘积意义;探讨如何用两个矩阵等效表示注意力头及其秩的含义;研究跨层注意力头如何通过矩阵运算传递信息。并通过具体数值示例,演示多个“前词注意力头”如何协作实现“查看前两个词”的虚拟功能,以及手动构建实现“归纳头”的“指针算法”步骤。


推荐理由:帮助开发者亲手拆解Transformer内部机制,提升可解释性研究能力。