Scene-OTC — Pairwise preference: leaderboard
Overall scene preference accuracy: 94 two-scene comparisons and 55 four-scene questions, with 2×94 and 4×55 evaluated position permutations reported by table. Four-scene accuracy requires the entire order correct. Percentages rounded by source. Candidate judges supplied rendered scenes; these are critique/ranking measurements, not scene generation. OTC-only and joint OTA+OTC SFT are distinct trained recipes, not base-model scores. Do not identify Scene-Critic with the separately released Scene-Builder checkpoint. Metric: Accuracy (%). Higher is better. Source: https://zju-omniai.github.io/Scene-Designer/
Metric: Accuracy (%). Source: zju-omniai.github.io. Saturation forecast: Around January 2027. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 72 | #109 |
| 2 | Qwen 3 VL 8B Instruct | 72 | #414 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #7 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=scene-otc-pairwise-preference · How It Works · Data refreshed daily, snapshot 2026-10-17.