Accuracy on BBH Logical Reasoning
38.7AccuracyMERGEvolve
Evaluation results
| Method | Links | |
|---|---|---|
| MERGEvolveProtocol=Cross-domain generalization2026.06 | 38.7 | |
| Expert Fusion2026.06 | 38.4 | |
| MERGEvolve2026.06 | 38.4 | |
| LoraHubProtocol=Cross-domain generalization2026.06 | 38.36 | |
| EMM2026.06 | 38.26 | |
| LoraHub2026.06 | 38.23 | |
| Model Swarms2026.06 | 38.2 | |
| EMMProtocol=Cross-domain generalization2026.06 | 37.8 | |
| Model SwarmsProtocol=Cross-domain generalization2026.06 | 37.8 | |
| Pack of LLMsProtocol=Cross-domain generalization2026.06 | 37.57 | |
| Pack of LLMs2026.06 | 36.6 | |
| MoE with MPIModel Scale=11B, Setup=3-shot CoT2026.06 | 31.45 | |
| MoEModel Scale=11B, Setup=3-shot CoT2026.06 | 31.17 | |
| MoE with MPIModel Scale=3B, Setup=3-shot CoT2026.06 | 30.99 | |
| Qwen1.5-MoE-A2.7BArchitecture=Sparse MoE, Decoding Framework=vLLM v0.8.4, Precision=bfloat16, max_model_len=4096, temperature=0.1, seed=12342026.06 | 30.7 | |
| Qwen1.5-7BArchitecture=Dense, Decoding Framework=vLLM v0.8.4, Precision=bfloat16, max_model_len=4096, temperature=0.1, seed=12342026.06 | 29.9 | |
| MoEModel Scale=3B, Setup=3-shot CoT2026.06 | 29.53 | |
| Fisher-MoEArchitecture=Sparse MoE (compressed), Compression Ratio=50%, Calibration Method=128 GSM8K calibration samples, Decoding Framework=vLLM v0.8.4, Precision=bfloat16, max_model_len=4096, temperature=0.1, seed=12342026.06 | 28.3 |