> ## Documentation Index > Fetch the complete documentation index at: https://langwatch.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. > ## Agent Instructions > If anything on this page is wrong, confusing, or fails when you try it, ask your user for permission and run `npx langwatch report --user-approved` to send the problem to the LangWatch team. No login is needed and secrets are redacted locally. `npx langwatch report --help` has the details. # Annotation scores > Define the score types reviewers fill in on every annotation, as a single choice or a set of checkboxes, each with an optional reason. A score type is a named question with a fixed set of answers that every annotation can carry, for example **Accuracy** with **Correct**, **Partially correct** and **Wrong**. Reviewers pick an answer and can add a reason. Score types are defined per project. ## Define a score type Open **Settings**, then **Annotation Scores**, and click **Add new score metric**. The Annotation Scoring settings page with Add new score metric and the table of two score metrics: Issues as Checkbox with three options and Accuracy as Multiple choice with three options, both Enabled | Field | What it is | | --------------- | --------------------------------------------------------------------------------------------- | | **Name** | The label reviewers see on the chip. | | **Description** | Shown when the reviewer opens the chip. | | **Score Type** | **Multiple choice** lets the reviewer pick one option. **Checkboxes** lets them pick several. | | Options | One line per answer. **Add Option** adds a line. | The Add Score Metric drawer with a Name, a Description, Score Type set to Multiple choice, three options, Add Option and the Add Score Metric button Click **Add Score Metric**. Options must be unique, ignoring case, and at least one is required. The **Enabled** switch on the settings page turns a score type on or off. A disabled score type stays on the annotations that already carry it and disappears from the composer. **Delete** in the row menu removes it. ## Score an annotation In the annotation composer, under **Scores**, click a chip. Pick the answer, type a reason in **Reason (optional)**, and click **OK**. A chip with a reason shows a message icon. See [Annotations](/docs/annotations/overview) for the composer. ## Where scores show up * The annotations inbox, in the folded **Scores** column and, through the **Columns** menu, in one column per score type. Hover over a score to read its reason. * The CSV export of the inbox, as `Name: value (reason)`. * The **annotations** source when you add a trace to a dataset: `score` and `score.reason` per score type, or all reviews as one text. * An annotation queue, which lists the score types its reviewers use. See [Annotation queues](/docs/annotations/queues). **Also check:** [Annotations](/docs/annotations/overview) for how a review is written, and [Datasets](/docs/datasets/overview) to store the scores with the rows.