Knowledge-based Language Understanding on MMLU-Pro (val)
42.9AccuracyEvoGM
Evaluation results
| Method | Links | |
|---|---|---|
| EvoGMBackbone=Qwen2.5-1.5B2026.05 | 42.9 | |
| Single BestBackbone=Qwen2.5-1.5B2026.05 | 38.6 | |
| CMABackbone=Qwen2.5-1.5B2026.05 | 38.6 | |
| PSO-MergingBackbone=Qwen2.5-1.5B2026.05 | 38.6 | |
| Model SwarmBackbone=Qwen2.5-1.5B2026.05 | 37.1 | |
| TIESBackbone=Qwen2.5-1.5B2026.05 | 32.9 | |
| Model SoupBackbone=Qwen2.5-1.5B2026.05 | 30 | |
| DAREBackbone=Qwen2.5-1.5B2026.05 | 30 | |
| MTLBackbone=Qwen2.5-1.5B2026.05 | 28.6 | |
| TABackbone=Qwen2.5-1.5B2026.05 | 28.6 | |
| BaseBackbone=Qwen2.5-1.5B2026.05 | 27.1 |