Predicate Grounding on Real Images Blocksworld
100F1 ScoreSymbolizer
Evaluation Results
| Method | Links | |
|---|---|---|
| SymbolizerModel=Gem.3.1-Pro2026.04 | 100 | |
| SymbolizerModel=Gem.3.1-FL2026.04 | 100 | |
| SYMBOLIZERBackbone=Gemini 3.1 Pro2026.04 | 100 | |
| SYMBOLIZERBackbone=Gemini 3.1 Flash Lite2026.04 | 100 | |
| ViLaInBackbone=Gem. 3.1-FL, Retry Setting=no retry2026.04 | 100 | |
| SYMBOLIZERBackbone=Gem.3.1-Pro2026.04 | 100 | |
| SYMBOLIZERBackbone=Gem.3.1-FL2026.04 | 100 | |
| SYMBOLIZERBackbone=Mistral Small2026.04 | 95.8 | |
| SymbolizerModel=Mistral Small2026.04 | 88.3 | |
| ViLaInBackbone=Gem. 3.1-FL, Retry Setting=two retries2026.04 | 80.2 | |
| SYMBOLIZERBackbone=Mistral Small2026.04 | 78.7 | |
| ViLaInBackbone=Gemini 3.1 Flash Lite, Retry Strategy=two retries2026.04 | 56 | |
| ViLaInBackbone=Gemini 3.1 Flash Lite, Retry Strategy=no retry2026.04 | 47.4 | |
| ViLaInModel=Gem. 3.1-FL, Retries=two retries2026.04 | 32.3 | |
| ViLaInModel=Gem. 3.1-FL, Retries=no retry2026.04 | 26.4 |