ChatVis_benchmark test cases
Test cases drawn from canonical visualization tasks, ParaView regression tests, and full scientific use cases are included in separate directories. Each test case includes any required input data, two versions of natural-language prompts, a ParaView Python script that satisfies the natural-language prompts, and a ground truth image generated by the script.
Two versions of prompts of differing verbosity describe the desired visualization. One prompt is written with detailed instructions and contains more visualization jargon that might appear in the ParaView documentation; we refer to this as the “full prompt.” The second prompt is written in a slightly relaxed conversational style and contains less visualization jargon; we refer to this as the “quick prompt.”
Each directory has a README with more details.