A paper is not an island. It stands on the shoulders of predecessors, and also treads on their scars. Dig backward to the root, then look forward — how the problem grew, what each person saw that others did not, and how solutions step by step approached the truth.
Core logic
The most common mistake in reading papers: looking only at the one in front of you, without knowing where it came from. The reverse-reading method does the opposite — first find whom this paper critiques and improves upon, then find whom that paper critiques, recursively for five levels, down to the source. Then turn around and read forward from the source.
After reading this way, what you gain is not the knowledge of one paper, but an understanding of an entire line of problem evolution.
Format constraints
Org-mode syntax
- Use
*bold*for bold (single asterisk),**bold** - Heading levels start from
*, no skipping levels
ASCII Art
All diagrams use pure ASCII characters. Allowed:+ - | / \ > < v ^ * = ~ . : # [ ] ( ) _ , ; ! ' " and spaces. Unicode drawing symbols forbidden.
Template authority
Output structure based on references/template.org。
Denote file specification
- Timestamp:
date +%Y%m%dT%H%M%S - Readable time:
date "+%Y-%m-%d %a %H:%M" - Filename:
{时间戳}--paper-river-{简短标题}__paper_river.org - Output directory:
~/Documents/notes/
Org file header
#+title: paper-river-{简短标题}
#+date: [{YYYY-MM-DD Day HH:MM}]
#+filetags: :paper:river:
#+identifier: {YYYYMMDDTHHMMSS}
#+source: {URL 或来源描述}
#+authors: {目标论文作者}
#+venue: {发表场所/年份}
Red lines
- Problem as the axis — The main thread of the whole article is "how the problem evolved", not "how the papers are arranged". Papers are supporting actors; the problem is the protagonist
- Spoken-language test — Would you tell a friend the history of a field's development this way? If not, revise
- Difference as the core — The focus of explaining each paper is "where it differs from the previous one", not introducing each paper independently
- Zero jargon — First land it in plain language, then mention the term in passing
- Unbroken logic chain — From the first paper to the last, the causal chain cannot break. Readers can feel "so that's why they did it this way"
- Honesty — If you can't find five levels, say how many you found. If the relationship between papers is uncertain, say it's uncertain. Don't fabricate citation relationships
Writing principles
- Difference-driven narrative — Don't write an independent summary of each paper and then stitch them together. Start each paragraph with "what problem this one saw in the previous one", letting the difference itself drive the narrative forward
- Transformation instead of definition — When explaining the difference between two approaches, continuously transform approach A into approach B. "If you remove X and add Y, you get Z" — ten times more powerful than "the difference between Z and X is..."
- Make reasoning explicit — Before each solution appears, first let readers feel the pressure of "not doing this is no longer possible". Simulate the process of discovery, not report the result of discovery
- One diagram is worth a thousand words — Draw a traceability map before the evolution narrative, and draw a compressed overview map after the narrative. Let readers have the full picture first, then enter the details, and return to the full picture after the details
Execution
1. Obtain the target paper
- arxiv URL → WebFetch
- PDF → Read (note the pages parameter limit)
- Paper title → WebSearch to find the full text
Make sure to obtain: title, authors, abstract, introduction (especially the critique of prior work in related work / introduction).
2. Extract the critique chain clues
Carefully read the target paper's introduction and related work section. Identify:
- Where it explicitly says "prior method X has problem Y"
- Which paper(s) it claims to improve upon
- Who the baselines it compares against are
From this, pin down the core papers being critiqued/improved upon(usually 1-3 papers; pick the most direct line).
3. Recursive tracing (deep research)
For the core prior papers found in step 2, repeat the same process: whom are they critiquing? Whom are they improving upon?
Recursion rules:
- Recurse at most 5 levels (stop at level 5 or at the foundational paper of the field)
- At each level, follow only the line most relevant to the problem, without diverging
- If at some level no clear target of critique can be found, stop there
Use the Research skill (deep research mode) to obtain key information for each level's papers. For each paper, obtain at least: title, authors, year, core problem, core solution, and critique points against prior work.
4. Frontier extension
In the opposite direction: after the target paper, are there new papers critiquing/improving upon it?
Also use the Research skill to search for:
- Follow-up work that cites the target paper
- The latest progress on the same problem
Find the 1-3 most relevant follow-up papers and obtain the same information.
5. Build the evolution line
Organize the results of steps 3 and 4 into a timeline:
[最老] Paper_0 → Paper_1 → ... → [目标论文] → [后续论文]
Label each arrow with: what problem the later work saw in the earlier one.
6. Forward Feynman narrative
Start from the oldest paper and narrate forward. The key: not introducing each paper independently, but connecting them along the evolution of the problem.
For each paper, cover three things (with emphasis on differences):
- What specific problem it saw in prior solutions (illustrate with an example or scenario)
- The core idea of its solution (explain clearly with an analogy)
- What new problem this solution left behind (naturally transitioning to the next paper)
7. Draw diagrams
Two diagrams:
- Tracing map: placed before the evolution narrative, showing the citation/critique relationships among the papers
- Problem-solution overview: placed after the narrative, compressing the whole line into one screen. Let people see at a glance how this line grew
8. Distill insights
After reading the whole line, answer:
- What change is really happening behind this evolution line? (not surface-level technical iteration, but a deeper cognitive shift)
- Where is it most likely to go next?
9. Check red lines + generate files
Scan the red lines one by one. Additional checks:
- Is the causal chain coherent—read all the "what problem it saw" statements together and see whether the logic holds
- Are the differences prominent—is each paper's focus on "how it differs from the previous ones"
Read references/template.org, write according to the Denote specification into ~/Documents/notes/。
Acceptance
- The problem is the protagonist: After reading, what you remember is "how the problem evolved," not "which papers there are"
- Causality is continuous: From the first paper to the last, every turning point has a "therefore"
- Differences are clear: Each paper's unique contribution can be stated clearly in one sentence
- Outsiders can follow: A smart person who does not understand this field can retell this line of evolution after reading
- The two figures can be understood independently: Without reading the main text, you can still grasp the gist just by looking at the figures
- Honest labeling: Which are confirmed citation relationships and which are inferred should be clearly marked