Compare papers
Compare reported methods and evidence across two to twenty papers.
Select 2–20 papers in the library and choose Compare. You can also start a Comparison from Threads and then pick papers. Choose the analysis types, then Run and save; each uses the project owner’s plan usage. Nothing is saved while you are still in the draft.
Choose the analysis you need
- Methodology: compare exact excerpts from stored methodology summaries and the papers’ own sections across research design, data, methods, evaluation and assumptions. A point shows a quote from the paper with its page when Distari finds it there word for word; NO QUOTE FOUND means the suggested quote wasn’t in the paper, so check that point against the source.
- Evidence table: extract task, dataset, method, metrics, result, sample size and limitations, then add your own columns. You can edit cells and export CSV or Markdown. A cell marked NOT REPORTED or UNVERIFIED · NO QUOTE FOUND offers Rerun free, which asks again for that cell only at no charge, once per cell. When the comparison also has a current Methodology result, a missed Dataset, Sample size, Method or Metrics cell shows what Methodology says about that paper, and Use this copies it into the cell.
- Research gaps: inspect leads supported by limitation or future-work sentences from at least two selected papers.
- Agreements and conflicts: inspect exact quoted claims from distinct papers and open the source evidence.
Methodology keeps incomplete descriptions incomplete. It does not add groups to reconcile a sample size or infer assumptions that the summaries do not report. A paper without a supported excerpt for a dimension shows Not reported in supplied summaries. If its stored summary has no method, a separate notice identifies the paper.
The completion caption says Based on stored summaries · verify against source papers. Open the linked papers to check the original methods: an excerpt can match a summary even when the summary needs correction. Older saved results keep their original content and charges; rerun Methodology to apply the current checks. Out-of-date results retain their historical warning.
A missing summary may need to be produced first and is quoted separately. A paper without extracted text can block a run. If the final quote is higher than the draft estimate, Distari asks again before running.
Saved comparisons live with questions in Threads. Adding or removing a paper marks completed results out of date. Their conclusions are limited to the selected papers; an optional outside-library search can offer further candidates, but it is not a systematic review or a novelty check.
Check for newer work outside the library
In the web app’s Research gaps view, select Check Semantic Scholar for newer work outside this library, then choose Run analysis or Rerun analysis and review the quote. Distari searches each supported gap theme. If no supported themes were found, it searches the selected papers’ topics using their titles instead.
The result reports newer candidates to review, no newer candidates found, or that the outside source was unavailable. Papers already in your project are excluded. No candidates is not proof that a gap is novel or that no related work exists. An older saved run may say no external-check outcome was recorded; run a new check to request one.
Review the cost of an empty rerun
An analysis can complete without findings that meet its evidence requirements and still use plan usage. On the web, repeating an empty result shows Your last run found nothing before you confirm, including if you change the Semantic Scholar option. The run can include missing summaries. Cancel leaves the saved result and your usage unchanged.
If the saved comparison list fails to load, follow Retry loading thread history to recover it without reloading the page.
Suggestions, rewrites, research gaps, methodology comparisons and summary translation use your plan’s usage on Free and Pro, and have no daily free allowance. In a shared project, the owner’s plan pays (see Plans and credits). Reading saved results is free; failed work releases the usage it held. A Research gaps or Agreements and conflicts run that finds nothing isn’t charged for the analysis, and the result says so; source summaries it wrote to prepare are still charged. Missing source summaries, comparison tables, agreement maps, drafting and full-text translation use plan usage too. Local PDF text extraction is free.