DeepPaperNote

single-paper deep reading workflow

A dense AI paper with objectives, attention maps, ablations, and claims

ALGORITHM · POLICY UPDATE

Attn(Q,K,V)=softmax( QK^T √d_k + B)V

𝓛(θ)=−∑_t=1^Tlog p_θ(y_t|x,y_<t)+λ KL(π_θ||π_ref)

sequence objective · policy gradient · regularized alignment

resolve

evidence

figures

synthesis

lint

Evidence before writing

Figure status stays explicit

Quality gates before save

domain

paper

status

A structured Obsidian note

核心信息

原文摘要翻译

创新点

方法主线

关键结果

Fig. 2 · training dynamics

Key formula

Â_i = r_i - mean(r)
∇_θJ = 𝔼[Â_i∇logπ_θ]

paper.md + images/

DeepPaperNote

把复杂论文整理成真正值得保留的 Obsidian 深度笔记。

Created by Huashu-Design