Paper brief

Keep a paper's method, evidence, novelty, and limitations.

Summarize the supplied academic paper in clear Markdown. Use these headings: Title and authors; Main goal and fundamental concept; Technical approach; Distinctive features; Experimental setup and results; Advantages and limitations; Conclusion.

Explain the research question, core idea, method, claimed novelty, experimental design, main findings, strengths, and limits. Prefer one concise paragraph per section. Define technical terms when needed, but retain the names of methods and models.

Preserve reported quantities, comparisons, uncertainty, and conditions. Separate the authors' claims from demonstrated results. Do not infer missing experiments, authors, or conclusions from an abstract. Mark unavailable information as not supplied. End with the main finding and the limitation that most affects its interpretation.

Example

Brief of an illustrative caching paper

Before 93 words

The paper "CacheLite" is by N. Chen. It asks whether reusing cached answers can reduce latency in a question-answering system. The method uses exact matching against a cache and is compared with a system that has no cache. The evaluation used 1,000 repeated questions; on that dataset, median latency fell by 30%. The authors describe the implementation as simple, but do not compare implementation complexity with other methods. They did not test new questions. The paper concludes that caching helped on this repeated-query test, while benefits on less repetitive workloads require further evaluation.

After 73 words

Title and authors CacheLite, N. Chen. Main goal and fundamental concept Reduce answer latency through reuse. Technical approach Compare exact-match caching with no cache. Distinctive features Authors claim simplicity; comparative complexity evidence is not supplied. Experimental setup and results Median latency fell 30% on 1,000 repeated questions. Advantages and limitations Faster on the tested workload; new questions were not tested. Conclusion Caching helped on repeated queries. Benefits on less repetitive workloads remain uncertain.

Examples illustrate the method. They do not measure model output.

Files and sources

prompt.txt Complete prompt as plain text.
SKILL.md The same prompt with a skill header.
llms.txt (JSON) Prompt, metadata, example, and sources.
LICENSE MIT license text.

Adapted from Fabric · summarize_paper. Source notes.