C3-Bench - Reversibility: leaderboard
Reversibility on C3-Bench (2026), a context-aware change-captioning benchmark: the percent of image pairs for which a GPT-5.2 judge finds that the before-to-after and after-to-before captions describe the same changes in opposite directions, averaged over three judge runs. Higher is better.
Metric: Reversibility (%): share of C3-Bench image pairs for which a GPT-5.2 judge finds that the model's before-to-after and after-to-before captions describe the same changes in opposite directions; mean of three judge runs; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 26 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 73 | #137 |
| 2 | GPT-5.1 | 63 | #229 |
| 3 | GPT-5.2 | 62 | #176 |
| 4 | Gemini 3 Flash | 62 | #164 |
| 5 | Gemini 2.5 Pro | 52 | #266 |
| 6 | GPT-4o | 49 | #741 |
| 7 | Gemini 2.5 Flash | 48 | #453 |
| 8 | Qwen 3 VL 32B Instruct | 47 | #585 |
| 9 | InternVL3-38B | 46 | #903 |
| 10 | GPT-5 Mini | 45 | #348 |
| 11 | InternVL3-78B | 45 | #738 |
| 12 | Llama 4 Scout Instruct | 44 | #1543 |
| 13 | Qwen 3 VL 8B Instruct | 40 | #981 |
| 14 | Qwen 2.5 VL 32B Instruct | 40 | #1053 |
| 15 | Qwen 2.5 VL 72B Instruct | 38 | #846 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=c3-bench-reversibility · How It Works · Data refreshed daily, snapshot 2026-11-13.