Complex Multimodal Reasoning on MM-Star
55.44Reasoning ScoreRecursiveVLM
Evaluation results
| Method | Links | |
|---|---|---|
| RecursiveVLMTraining Dataset=Data2, Train Steps=2, Eval Step=22026.02 | 55.44 | |
| RecursiveVLMTraining Dataset=Data2, Train Steps=2, Eval Step=12026.02 | 52.16 | |
| RecursiveVLMTraining Dataset=Data1, Train Steps=2, Eval Step=22026.02 | 52.1 | |
| RecursiveVLMTraining Dataset=Data1, Train Steps=2, Eval Step=12026.02 | 51.24 | |
| Standard TransformerTraining Dataset=Data2, Train Steps=1, Eval Step=12026.02 | 50.14 | |
| Standard TransformerTraining Dataset=Data1, Train Steps=1, Eval Step=12026.02 | 45.03 | |
| Vanilla recursionTraining Dataset=Data1, Train Steps=2, Eval Step=12026.02 | 42.38 | |
| Vanilla recursion w/ RMSNormTraining Dataset=Data1, Train Steps=2, Eval Step=12026.02 | 41.7 | |
| Vanilla recursionTraining Dataset=Data1, Train Steps=2, Eval Step=22026.02 | 41.47 | |
| Vanilla recursion w/ RMSNormTraining Dataset=Data1, Train Steps=2, Eval Step=22026.02 | 40.28 |