SDGBiasBench - MCQ (Full Context Values): leaderboard
Three-option ordinal questions about Sustainable Development Goal indicators of a place from satellite and street-level imagery, with all structured SDG variables given with their values; SDGBiasBench (Lin et al., 2026). Accuracy in percent; higher is better.
Metric: Accuracy (%; three-option ordinal questions on SDG indicators from satellite and street-level imagery, all structured SDG variables given with their values). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.0 Flash | 62 | #702 |
| 2 | GPT-4o | 61 | #741 |
| 3 | InternVL3-14B | 61 | #1184 |
| 4 | InternVL3-8B | 59 | #1695 |
| 5 | Claude 3.5 Sonnet | 56 | #747 |
| 6 | Qwen 2.5 VL 7B | 54 | #1796 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=sdgbiasbench-mcq-full-context-values · How It Works · Data refreshed daily, snapshot 2026-11-12.