MergeBench
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Multi-task Language Modeling | MergeBench | Instruction Score39.56 | 21 | |
| Model Merging Evaluation | MergeBench | MATH-500 Score52.6 | 12 | |
| Generative Language Model Merging | MergeBench Expert models Llama-3.2-3B (test) | Instruction Score53.52 | 11 | |
| Multi-task Language Understanding | MergeBench Llama-3.2-3B (test) | Average Score44.1 | 11 | |
| Vision-Language Multi-task Performance | MergeBench (Vision-Language tasks: MMSI-Bench, EmbSpatial, MMMU_Med, PathVQA, OCRBench, CharXiv) | MMSI-Bench32.6 | 11 |
Showing 5 of 5 rows