论文倒读法：给一篇论文，递归找出它批判和改进的前序论文（最多5层），再找它之后的最新进展，从源头正向讲述问题演化史。以问题为轴，费曼式讲解每篇论文看到的问题和解法创新。Use when user shares a paper and wants to understand its intellectual lineage, citation chain, problem evolution, or says '倒读', '论文溯源', '论文脉络', 'paper river', 'paper connects', 'trace back', '这篇论文的来龙去脉', '论文演化'. Also trigger when user wants to understand how a research problem evolved across multiple papers.
ljg-paper-river is an AI research analysis skill designed to reconstruct the intellectual lineage of an academic paper through a recursive “reverse reading” methodology. Instead of treating a paper as an isolated artifact, the skill traces backward through the papers it criticizes, improves upon, or directly builds on, recursively following the most relevant problem lineage for up to five layers. After reaching the conceptual origin of the problem, it reverses direction and narrates the forward evolution of the research area, showing how each generation of papers identified weaknesses in prior approaches and introduced new solutions. The skill focuses on problem evolution rather than paper summaries, helping users understand how ideas emerged, transformed, and accumulated over time.
The skill extracts signals from introductions, related work sections, baselines, and explicit criticism statements in papers to identify the core citation chain. It builds a structured timeline connecting foundational work, the target paper, and later developments that extend or challenge it. The output emphasizes causal reasoning, difference-driven storytelling, and Feynman-style explanation techniques, translating technical progress into intuitive narratives. It also generates structured ASCII diagrams such as citation lineage maps and compressed problem-solution overviews. The workflow includes recursive paper tracing, forward-looking literature extension, timeline construction, comparative analysis, insight synthesis, and generation of Denote-compatible Org-mode notes.
This skill is intended for researchers, graduate students, independent learners, and technical readers who want to deeply understand how a research problem evolved across multiple papers rather than only reading isolated publications. It is especially useful for literature reviews, research onboarding, academic synthesis, paper lineage tracing, and understanding conceptual shifts within a field. Users can invoke it when analyzing a specific paper, investigating research origins, studying citation chains, or exploring how competing solutions evolved over time.
The skill supports arXiv URLs, PDF papers, or paper titles that can be searched online. It requires access to the paper’s abstract, introduction, and related work sections to identify citation and criticism chains.
The recursive backward tracing process goes up to five layers deep or stops earlier if no clear predecessor or criticism chain can be identified.
No. The skill is designed around problem evolution and comparative differences between papers rather than isolated summaries. Each paper is explained in relation to the weaknesses or limitations of the previous work.
The skill generates structured Org-mode documentation following Denote naming conventions, including metadata headers, ASCII-based diagrams, evolution timelines, and narrative analysis.
The skill explicitly avoids fabricating relationships. If a citation chain or criticism link is uncertain, it is labeled as uncertain rather than presented as confirmed.
Quick Setup:
.claude/skills/Repository
lijigang/ljg-skills