https://drive.google.com/file/d/1BU5bV3X5w65DwSMapKcsr0ZvrMRU_Nbi/view
Do large language models need sensory grounding for meaning and understanding?
Spoiler: YES!
Yann LeCun
Courant Institute & Center for Data Science, NYU
Meta – Fundamental AI Research
2023-03-24
Do large language models need sensory grounding for meaning and understanding?
Spoiler: YES!
Yann LeCun
Courant Institute & Center for Data Science, NYU
Meta – Fundamental AI Research
2023-03-24
Vintage NLP for Work
https://drive.google.com/file/d/1BU5bV3X5w65DwSMapKcsr0ZvrMRU_Nbi/view Do large language models need sensory grounding for meaning and understanding? Spoiler: YES! Yann LeCun Courant Institute & Center for Data Science, NYU Meta – Fundamental AI Research…
vxTwitter / fixvx
💖 86 🔁 11
💖 86 🔁 11
Sander Dieleman (@sedielem)
Some thoughts on non-AR language models, and what it might take to dethrone autoregression: https://sander.ai/2023/01/09/diffusion-language.html
https://yimeixiang.wordpress.com/teaching/ 一点逻辑语义学课程材料。没有太争议的主题，如语义的本体论问题、语义语用界面等，偏具体的语义学技术，对入门友好。
Yimei Xiang ∙ 向伊梅
Teaching
Semantics II (Graduate, Rutgers, Spring 2021) Unit 1: Compositionality and Binding Part I: Compositionality, Heim & Kratzer theory of pronoun binding Readings: Heim & Kratzer (1998: Chapter…
Binding Language Models in Symbolic Languages
https://lm-code-binder.github.io/
ICLR23 top-25% 终于也开始实际地评估LLM转换形式语言的效果了
https://lm-code-binder.github.io/
ICLR23 top-25% 终于也开始实际地评估LLM转换形式语言的效果了
https://storage.googleapis.com/deepmind-media/Era-of-Experience%20/The%20Era%20of%20Experience%20Paper.pdf
https://lilianweng.github.io/posts/2025-05-01-thinking/
https://ysymyth.github.io/The-Second-Half/
三位RL学者近期巨作：转向经验学习 (R. Sutton) — 思考如何思考 (Lilian Weng) — 算法让位给产品(Shunyu Yao)
欢迎讨论～
https://lilianweng.github.io/posts/2025-05-01-thinking/
https://ysymyth.github.io/The-Second-Half/
三位RL学者近期巨作：转向经验学习 (R. Sutton) — 思考如何思考 (Lilian Weng) — 算法让位给产品(Shunyu Yao)
欢迎讨论～
What, How, Where, and How Well? A Survey on Test-Time Scaling in Large Language Models
https://testtimescaling.github.io/
主要是这图整挺好（
另附一篇相似目标的综述，称为长CoT
https://long-cot.github.io/
个人感觉这两篇只可以作为检索工具使用，或者帮新入学研究生建立低分辨率地图。推理发展内驱力还并不能完整地被这个分类学刻画，在机理解释性不足的现在也没有条件做理论家的工作，有那功夫直接做研究文章了（
https://testtimescaling.github.io/
主要是这图整挺好（
另附一篇相似目标的综述，称为长CoT
https://long-cot.github.io/
个人感觉这两篇只可以作为检索工具使用，或者帮新入学研究生建立低分辨率地图。推理发展内驱力还并不能完整地被这个分类学刻画，在机理解释性不足的现在也没有条件做理论家的工作，有那功夫直接做研究文章了（
Forwarded from A Math Book
Theory of Deep Learning.pdf
6.2 MB
Theory of Deep Learning ( Sanjeev Arora ). 2022
Forwarded from Parallel Experiments (Linghao Zhang)
https://gregorygundersen.com/blog/2025/10/01/large-language-models/
预感这篇会是 LLM Researcher 必读：作者把跨越数十年的语言模型研究梳理成了一条清晰的时间线，讲述我们是怎么一步一步得到今天的 transformer based LLM 的。文章的思路非常 from first principles，并且用前后一致的符号串起了 N 篇不同的论文的要点。
非常喜欢文尾的一段话：
> If you feel that it’s a bit perverse that next-word prediction is a sufficient objective to solve elite math problems, if this feels like a stochastic parrot outsmarting you, then you might feel some of the discomfort early linguists felt at statistical language modeling. This is the visceral feeling of the bitter lesson. Our specialized knowledge feels expendable and our intuitions about understanding seem irrelevant in the face of raw computation and speed.
预感这篇会是 LLM Researcher 必读：作者把跨越数十年的语言模型研究梳理成了一条清晰的时间线，讲述我们是怎么一步一步得到今天的 transformer based LLM 的。文章的思路非常 from first principles，并且用前后一致的符号串起了 N 篇不同的论文的要点。
非常喜欢文尾的一段话：
> If you feel that it’s a bit perverse that next-word prediction is a sufficient objective to solve elite math problems, if this feels like a stochastic parrot outsmarting you, then you might feel some of the discomfort early linguists felt at statistical language modeling. This is the visceral feeling of the bitter lesson. Our specialized knowledge feels expendable and our intuitions about understanding seem irrelevant in the face of raw computation and speed.
Gregorygundersen
A History of Large Language Models
Parallel Experiments
https://gregorygundersen.com/blog/2025/10/01/large-language-models/ 预感这篇会是 LLM Researcher 必读：作者把跨越数十年的语言模型研究梳理成了一条清晰的时间线，讲述我们是怎么一步一步得到今天的 transformer based LLM 的。文章的思路非常 from first principles，并且用前后一致的符号串起了 N 篇不同的论文的要点。 非常喜欢文尾的一段话： > If you feel that it’s…
惭愧第一次看到 hinton, 1986，整理下可能至少留意如下几种“分布的”的语源：
- distributional semantics/hypothesis/representation (Harris, 1956)
: 指用语料中分布情况来定义成“语义”，在当时可能希望把“分布”数学化但其实还是挺语言学的
- distributed representation(Hinton, 1986): 指特征并非是组织好的而是像神经元一样散落地在各处激活
而无论如何我们都可以用更数学（至少更概率统计）的方式来描述更当代的 representation了，上述语源可能也从 word2vec 开始合流，不再有区分的必要
- distributional semantics/hypothesis/representation (Harris, 1956)
: 指用语料中分布情况来定义成“语义”，在当时可能希望把“分布”数学化但其实还是挺语言学的
- distributed representation(Hinton, 1986): 指特征并非是组织好的而是像神经元一样散落地在各处激活
而无论如何我们都可以用更数学（至少更概率统计）的方式来描述更当代的 representation了，上述语源可能也从 word2vec 开始合流，不再有区分的必要