A paper = one increment. Existing research has reached a certain boundary, and this paper claims to push one step further. Your task: where exactly does this step land, and is it solid.
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 are forbidden.
Denote file conventions
- Timestamp:
date +%Y%m%dT%H%M%S - Human-readable time:
date "+%Y-%m-%d %a %H:%M" - Filename:
{时间戳}--paper-{简短标题}__paper.org - Output directory:
~/Documents/notes/
Org file header
#+title: paper-{简短标题}
#+date: [{YYYY-MM-DD Day HH:MM}]
#+filetags: :paper:
#+identifier: {YYYYMMDDTHHMMSS}
#+source: {URL 或来源描述}
#+authors: {作者列表}
#+venue: {发表场所/年份}
After writing the file, report the path.
Execution
1. Obtain the content
- arxiv URL → ljg-fetch or WebFetch
- PDF → Read (note the pages parameter limit)
- Local file → Read
- Paper name → WebSearch
Make sure you get: title, authors, abstract, introduction, method, experiments/results, conclusion.
2. Split
A paper has a fixed skeleton, but beneath the skeleton hides the real structure:
- Gap: How far has existing research gotten? What has not been achieved yet? Which seam does this paper fill?
- Assumptions: Explicit and implicit assumptions.
- Method: What method is used to fill the seam? Grasp the verb structure—what operation is it performing?
- Evidence: What data/experiments prove the method holds up? Main results.
- Contribution claims: What the authors themselves say they contributed.
3. Squeeze the increment
Before this paper vs after this paper, what has the world gained?
- One sentence: What does this paper let us know that we did not know before?
- Core mechanism diagram: Use an ASCII diagram to draw the internal structure of the method—components, data flow, operational relationships. After seeing it, one knows "how it turns internally."
4. Plain-language method
Find an everyday analogy for the core mechanism, so that readers can retell the method logic after following it through.
- Grasp the verb structure of the method: what acts on what? What flows to what?
- Find something in everyday experience with the same structure. The analogy must bear weight—every key component of the method can be mapped to a part of the analogy.
- Walk through the method from beginning to end along the analogy.
The analogy follows immediately after the ASCII diagram, as a continuation of the same paragraph.
5. Key concepts
1-3 key concepts in the paper for understanding the whole text. Explain using the Feynman technique:
- Assume the reader knows nothing, start from the most basic place
- Give a concrete example for each concept
- After reading, the reader can retell it in their own words
6. Napkin sketch
Pick up a napkin in a cafe and draw for a friend: "People used to think this way; this paper says we should think this way."
- Identify the mainstream framework and this paper's framework
- Use ASCII diagrams side by side or one above the other, annotating key structural differences
- The sentence below the figure points out: from X to Y, what is the core change
Draw a structural comparison; the more concise the better.
7. PhD advisor review
Switch identity: a PhD advisor who has supervised graduate students in this direction for twenty years. A student comes to you with a paper, and you are judging whether this thing is worth taking seriously. The paper tests the system—taste in topic selection, maturity of method, sincerity of experiments, writing craft.
In plain words, like chatting with a student in the office:
- Taste in topic selection: Is the problem worth doing? A real gap or a manufactured gap?
- Maturity of method: Cleverness or brute force? Is there a more natural approach that was overlooked?
- Sincerity of experiments: Are the baselines fair? Is the ablation thorough? Do the numbers withstand questioning?
- Writing craft: Is there any laziness where things should be clearest?
- Verdict: strong accept / weak accept / borderline / weak reject / strong reject, with a one-sentence reason.
Say what is good is good, and say where what is bad falls short.
8. Inspiration
After finishing the paper, ask: what use is it to me? Probe in three ways:
- Transfer: Can some mechanism/perspective from the paper be directly transplanted to upgrade some component in my system?
- Mix and match: Can some component of the paper be combined with something I already have to produce something new that neither has alone?
- Inversion: Is the paper's approach contrary to my default assumptions? What blind spot does it reveal?
Annotate the types that hit and expand them. Skip the ones that do not hit. If none of the three hit, say "none".
9. Stitching and hygiene
According to references/template.org structure, weave the above results into a coherent analysis.
After writing, scan through:
- No more than two negative parallel constructions in the whole text
- Change three-part constructions to two or four items
- No more than two em dashes in the same paragraph
Make a revision checklist and generate the file after confirmation.
10. Generate Org file
Get the timestamp according to the Denote convention, read references/template.org, write to ~/Documents/notes/。
Quality standards
- The increment must be sharp: State before vs after in one sentence. If you cannot, you have not found it.
- The sketch must be clear at a glance: Without reading the main text, one can still roughly understand the direction of change.
- The PhD advisor must be like a PhD advisor: Have judgment and tact, with a verdict in the final sentence.
- Zero sense of fragmentation: After reading, it feels like one person telling you, "I read a paper; what it did, whether it is good, and what use it is to me."