For fifty combinations of city, policy goal and target year, propose a package of transport measures that would reach the goal.
Multiple cities · LLM judge rating / 100 ↑ · 50 items
Recomputed from saved answers. Model labels retain the original version and repeat protocol. No overall model average.
| Model | Raw result | Valid / asked |
|---|---|---|
| GPT-5.5 | 82.964 | 50 / 50 |
| Gemini 2.5 Pro | 73.784 | 50 / 50 |
| Kimi 2.6 | 70.532 | 50 / 50 |
| Gemma-4-26B (API, temp 0) | 55.024 | 50 / 50 |
| llama (API, temp 0) | … | 0 / 50 |
| gptoss (API, temp 0) | … | 0 / 50 |
| glm53 (API, temp 0) | … | 0 / 50 |
| minimax3 (API, temp 0) | … | 0 / 50 |
| qwen (API, temp 0) | … | 0 / 50 |
| deepseek (API, temp 0) | … | 0 / 50 |
Correct answer known from: There is no single right answer here. Each proposal is read by an LLM judge and scored on five axes: effectiveness, cost, political acceptance, local specificity and local correctness.
A judged score is weaker evidence than a measured one. It says a reader found the answer plausible, not that it was right.
Click any cell to load its prompt and responses into the browser below.
Pick a scenario here, or click a cell in the heat-map above.
Each model receives the exact same prompt above.