Contact

For questions about the dataset, scoring methodology, model coverage, or published model-output profiles, contact Jonathan R. Reed through the author site.

Useful notes include the page URL, model slug, run ID, axis name, artifact link, and the exact claim that needs review. That keeps feedback attached to the evidence surface instead of turning it into a broad argument about model politics.

Researchers interested in using PoliBench data, extending the question bank, or collaborating on AI political-response analysis are welcome to reach out.

Collaboration questions are most useful when they name the intended use: replication, new axes, external anchors, human baseline collection, validation design, or a paper-facing critique of the current release boundaries.

For privacy questions, data review requests, or removal inquiries related to published benchmark rows, see the Privacy Policy or contact Jonathan R. Reed.

Can I use PoliBench data in my research? Live model-output data is browsable on this site and can be referenced in research. Frozen paper packs are not currently republished. Please cite the release and preserve the stated evidence level when referencing scores or model-output profiles.

How often are models re-evaluated? New model routes are considered through the locked release process. Paper-facing updates require preflight, canary, audit, and artifact checks before publication.

Can I suggest new questions or topics? The question bank is versioned and evolves over time. Proposals for new axes, topic expansions, or wording refinements are welcome for consideration in future releases.