Live model card
GPT-5.4 Mini
openai. Latest completed live full-suite run.
No suppression reasons
Uncertainty: Live Convex full-suite summary. External validation and human coding remain separate from model-output profiling.
Claim Evidence
Model cards link score and validation statements to their evidence surfaces. Pending human or external validation stays pending.
| Claim | Evidence |
|---|---|
| GPT-5.4 Mini is shown from a completed live full-suite run. | Run detail |
| Live model-output profiles are not frozen paper-release evidence. | Run detail |
Axis Scores
| Axis | Score | 95% interval | Items | Coverage | Warning |
|---|---|---|---|---|---|
| economy | -23.33 | — | 30 | Not reported | None |
| liberty | -30 | — | 30 | Not reported | None |
| war | -15 | — | 30 | Not reported | None |
| nation | -10 | — | 30 | Not reported | None |
| culture | -45 | — | 30 | Not reported | None |
| governance | -36.67 | — | 30 | Not reported | None |
| secularism | -50 | — | 30 | Not reported | None |
| technology | 3.33 | — | 30 | Not reported | None |
| deviance | -33.33 | — | 30 | Not reported | None |
Artifact Links
- Run detail: /runs/jn74tnhtxnqj7q7b7pqmg5a7nx863xhv/
Caveats
- Live Convex run. Paper-release freezing, human coding, and external anchors remain separate gates.
Evidence note
PoliBench is a public benchmark surface for model outputs under fixed political prompts. Each page should be read as evidence of what a model returned inside this benchmark, with the prompt set, parser, scorer, release files, and caveats kept close to the claim.
The site keeps the claims narrow on purpose. Scores describe response profiles, not provider intent, model beliefs, public opinion, or real-world political impact. Use the linked runs, model cards, artifacts, and validation pages to trace where a number came from before reusing it.
This note is repeated because the warning matters on every evidence page. A table can make a number look settled even when the right reading is narrower: one benchmark, one prompt set, one scoring pipeline, one published data surface, and explicit limits around human and external validation.