-
Reading Notes of "The Origin of Concepts"
The Origin of Concepts 阅读笔记
-
Reading Notes: Attention Sinks in Language Models
从 repeated token、over-mixing 和几何视角理解 attention sink，并讨论 diffusion language models。
-
Reading Notes: Layer Normalization in Transformers
关于 Pre-LN、Post-LN、训练稳定性和 Transformer 架构改进的阅读笔记。
-
Reading Notes: Latent Reasoning and Reinforcement Learning
关于 latent reasoning、test-time adaptation 和 reinforcement learning 的论文阅读笔记。
-
Reading Notes: Latent Chain-of-Thought and Recurrent Computation
从 discrete and continuous tokens 到 recurrent computation，整理 latent reasoning 的一些方法。
-
Physics of Language Models 阅读笔记
在ICML听了Allen Zhu的tutorial，觉得很有启发，最近发现phys of LMs系列还在不断更新，写一篇阅读笔记系统整理一下。
-
大模型机制可解释性tutorial
整理一下language model这块的入门资料。
-
调研 reasoning models 的推理机制
a blog about reasoning mechanisms of reasoning models