Image Captioning on Captioning Benchmarks 1.0 (test)
67.8DCScoreGemini-1.5-Pro
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Gemini-1.5-ProCategory=Proprietary MLLMs, Arena Length=3622026.05 | 67.8 | 90 | 79.8 | 73 | |
| GPT-4-TurboCategory=Proprietary MLLMs, Arena Length=2202026.05 | 66.5 | 89.2 | 82.2 | 70.9 | |
| Gemini-1.5-FlashCategory=Proprietary MLLMs, Arena Length=1972026.05 | 59.9 | 83.9 | 81.2 | 65.6 | |
| BalCapRLBackbone=QwenVL2.5-7B, Arena Length=1922026.05 | 53.4 | 79.1 | 28.5 | 58.7 | |
| BalCapRLBackbone=QwenVL2.5-3B, Arena Length=1752026.05 | 50.8 | 75 | -3.8 | 49.3 | |
| RubiCap-7BBackbone=QwenVL2.5-7B, Arena Length=1762026.05 | 50.5 | 76 | 22.7 | 53.9 | |
| CapRL-3BBackbone=QwenVL2.5-3B, Arena Length=4032026.05 | 48.6 | 82.6 | -50.6 | 47.4 | |
| BaselineBackbone=QwenVL2.5-7B, Arena Length=1362026.05 | 46 | 74.9 | 13.7 | 51.1 | |
| GPT-4oCategory=Proprietary MLLMs, Arena Length=832026.05 | 45.8 | 78.7 | 10.3 | 56.1 | |
| BaselineBackbone=QwenVL2.5-3B, Arena Length=1312026.05 | 43.3 | 70 | -34 | 46.2 | |
| RubiCap-3BBackbone=QwenVL2.5-3B, Arena Length=1492026.05 | 43 | 71.1 | -29.5 | 46.8 | |
| BalCapRLBackbone=LLaVA-1.5-7B, Arena Length=1242026.05 | 36.6 | 55.4 | -65 | 43.4 | |
| FEEDQUILLBackbone=LLaVA-1.5-7B2026.05 | 34.5 | — | — | — | |
| BaselineBackbone=LLaVA-1.5-7B, Arena Length=742026.05 | 23 | 46.4 | -94 | 26.9 |