extract-paper-tablelisted

# Extract Paper Table Turn a pile of papers into a clean table: **one row per paper, one column per attribute.** This is the data-extraction step of a literature review or systematic review, done consistently across the whole set. ## Inputs - **Papers** (up to ~30). For each: title, authors/year, abstract, and full text or a long excerpt if available. The more text, the better the extraction. - **Columns.** Either use the defaults or define your own. A column is a `name` plus an optional `hint` telling the extractor exactly what to pull. ### Default columns | Column | Hint | |---|---| | Sample | Study population: who/what was studied, with size if available (e.g. "n=128 university students", "mice", "open-source repos"). | | Method | Research method or design (e.g. "RCT", "qualitative interviews", "meta-analysis of 14 trials", "regression on panel data"). | | Setting | Where/when conducted (e.g. "UK NHS hospitals 2019–2021", "simulation in PyTorch", "cross-country panel 1990–2020"). | | Key finding | The single most important quantitative/qualitative finding the authors emphasise. Concrete, with numbers where possible. | | Limitation | Author-acknowledged or obvious limitation: small n, narrow context, observational only, etc. | ## Method — extract one row at a time **Process each paper independently, one row per paper** (not one cell at a time, not the whole table in one shot). Per-row extraction lets the model see all columns together — it knows "sample size" and