​

Table

All results are displayed in the results tab in the table.

It’s possible to filter the results by the model ran, the current status, who it was ran by, which test suite it was from, as well as to search by run name or test suite name. You can show and hide columns of the table by pressing “Show Columns”, and also sort by pass rate and when it was started.

There are a few actions available from this view:

• Compare: Compare the results of two runs. Note: the two runs must be from the same test suite.
• Archive: Archive a run. This means it will no longer be displayed in the list runs by default (although you can still see it by turning on “Show Archived Runs”)
• Export: Export one or more runs to a CSV file.

​

Viewing a Single Run

Clicking on a row will bring you to the single run page.

On the left, it shows the results of every test - including the input, the output from the LLM, how it performed on each check, and other information. Ther right contains top-level information about the run - including statistics about overall performance, a free-text summary, and the parameters of the run.

​

Filtering

It’s possible to filter the results to only view certain tests. First, you can filter based on failures - showing only the tests where some, all, or none of the checks failed.

It is also possible to search for a given string in the test. This searches over the input, output, checks, LLM feedback, and context by default.

Additionally, you can filter to only tests that have a given tag.

Finally, you can filter to only the tests there the model had an error when producing the output (e.g., if you were looking for cases where the model exceeded its token limit).

​

Confidence

By default, the Vals system will compute a confidence for every test - either “High” or “Low”. If the confidence is “High”, this means our system has flagged that it is very likely we’ve graded this output correctly. If the confidence is “Low”, then it means there is greater uncertainty or ambiguity in either the grading or in the criteria.

The confidence score is listed in each check next to the grade.

​

Statistics

Several statistics are reported for every run.

• Success Rate: This is the percentage of tests in which all tests passed. For example, lets say there are four tests, each with two checks each. The first two get 2/2 points, the third gets 1/2 points, and the last gets 0/2. The success rate is 25% - 2/4 tests had all their checks pass.
• Pass Rate: This is the raw percentage of checks that passed - regardless of which test they belong to. So, if there are 8 checks across 4 tests, and 6 passed, then the pass rate is 75%.
• Low Confidence Rate: This is the percentage of tests that the system flagged as low confidence.

Additionally, each check is by default given a certain category: e.g. “Correctness”, “Format”, “Style”, etc. The run result page will also show a pass percentage for each category (note: the categories can be overridden).

Finally, if tags are assigned to each test, it will show a performance breakdown by tag.

​

Comparisons

You can choose to compare two runs either through the table or when looking at a single run. This allows you to see the results of two runs side by side - this is commonly done to see the differences between two models.

We automatically compute several statistics, such as the likelihood of a statistically significant difference between the two runs.

You can also filter to only show tests that were marked differently by the auto-grader by pressing “Hide tests with no differences”.