paper-summary

Provide the paper's title or URL and the skill reads the paper and produces a structured summary covering bibliography, problem setting, contributions, method key ideas, evaluation, discussion, and related papers. The output follows a fixed format in concise Japanese.

The summary includes bibliographic information, problem setting, improvements over prior work, core technical ideas, an evaluation table with metrics and baselines, discussion points, related papers with one-line summaries, and detailed axis-based comparison tables with prior research.

Yes, you can ask questions about the paper's content after the summary. Answers are strictly based on what the paper states, and the skill will not invent information that is not written in the paper.

Yes, providing an arXiv URL or the paper title is sufficient to trigger the summary. The bibliographic section records the arXiv link or other publication source, along with venue, year, authors, and official implementation if available.

The summary quality depends on access to the paper's full text, and paywalled papers may not be fully readable. Numerical results are reported as trends rather than exact values, so consult the original paper for precise figures.