> ## Documentation index
> Fetch the complete documentation index at: [llms.txt](../llms.txt)
> Use this file to discover all available pages before exploring further.

# Check a dataset with Review mode

See only the cells worth a second look, check their sources, and mark them reviewed.

[Open a dataset](https://app.sleuth.studio)

You don't have to check every cell. Values with strong sources that agree score high, and Review mode puts the rest in front of you, with the reason.

## Review a dataset

1. **Open the dataset** and switch from the dataset view to **Review mode**.
2. **Filter** with the flags: **Low confidence**, **Inferred**, and **Sources disagree**. Start with low confidence.
3. **Open a cell** to see its score, Sleuth's explanation, and its sources. Sources open inside Sleuth: papers, press releases, earnings transcripts, and broker research.
4. **Decide.** Mark the cell reviewed, flag it, leave a comment, or edit it by double-clicking. To have Sleuth try again, re-run the cell with guidance; each re-run uses one credit.

## When sources disagree

For a disputed cell, Sleuth explains the conflict and suggests a value. Apply the suggestion or keep the current value. A common case: China's trial registry shows a Phase 3 start while ClinicalTrials.gov still lists Phase 2, and Sleuth used the newer record, which scored higher.

## Things to know

- Your edits and review marks are saved with the dataset, and edited values get a confidence of 100.
- You can undo a review mark.
- You can't edit a past version of a dataset, or a dataset someone else owns.

## Related

- [Confidence scores and labels](confidence.md): How Sleuth scores every value, and what sourced, inferred, and unknown mean.
- [Datasets: rows, columns, and cells](datasets.md): What a Sleuth dataset is, which cells use credits, and how to change one.
- [Run your first analysis](first-analysis.md): From sign-in to a finished analysis you can check and export, in five steps.

*Updated November 9, 2026.*
