Multi-domain reasoning on BBH
87.39AccuracyALRR
Evaluation Results
| Method | Links | |
|---|---|---|
| ALRRModel=Qwen3-4B2026.06 | 87.39 | |
| ALRR+ERSSModel=Qwen3-4B2026.06 | 87.29 | |
| LRSModel=Qwen3-4B2026.06 | 87.2 | |
| ERSSModel=Qwen3-4B2026.06 | 87.18 | |
| ALRR+LRSModel=Qwen3-4B2026.06 | 87.17 | |
| phi-balancingModel=Moonlight-16B-A3B-Instruct2026.05 | 85.74 | |
| LRSModel=Phi-42026.06 | 85.61 | |
| ALRRModel=Phi-42026.06 | 85.56 | |
| ERSS+LRSModel=Qwen3-4B2026.06 | 84.81 | |
| BaselineModel=Qwen3-4B2026.06 | 84.78 | |
| BaselineModel=Phi-42026.06 | 84.73 | |
| ERSSModel=Phi-42026.06 | 84.67 | |
| ST-MoEModel=Moonlight-16B-A3B-Instruct2026.05 | 82 | |
| ERSSModel=Gemma3-12B-it2026.06 | 77.38 | |
| ALRRModel=Gemma3-12B-it2026.06 | 77.19 | |
| LRSModel=Gemma3-12B-it2026.06 | 76.79 | |
| BaselineModel=Gemma3-12B-it2026.06 | 76.74 | |
| LRSModel=R1-Distilled-LLaMA-8B2026.06 | 74.14 | |
| phi-balancingModel=DeepSeek-MoE-Chat2026.05 | 73.92 | |
| ERSSModel=Phi-4-Reasoning2026.06 | 73.91 | |
| ALRRModel=R1-Distilled-LLaMA-8B2026.06 | 72.53 | |
| ALRR+ERSSModel=R1-Distilled-LLaMA-8B2026.06 | 72.03 | |
| ALRR+LRSModel=R1-Distilled-LLaMA-8B2026.06 | 72.03 | |
| ALRRModel=Phi-4-Reasoning2026.06 | 71.69 | |
| LRSModel=Phi-4-Reasoning2026.06 | 71.27 | |
| BaselineModel=Phi-4-Reasoning2026.06 | 71.23 | |
| ERSSModel=R1-Distilled-LLaMA-8B2026.06 | 70.9 | |
| BaselineModel=R1-Distilled-LLaMA-8B2026.06 | 70.48 | |
| ALRRModel=Gemma3-4B-it2026.06 | 69.92 | |
| ST-MoEModel=DeepSeek-MoE-Chat2026.05 | 69.86 | |
| ERSSModel=Gemma3-4B-it2026.06 | 69.67 | |
| LRSModel=Gemma3-4B-it2026.06 | 69.52 | |
| ERSS+LRSModel=R1-Distilled-LLaMA-8B2026.06 | 69.44 | |
| BaselineModel=Gemma3-4B-it2026.06 | 69.03 | |
| phi-balancingModel=DeepSeek-V2-Lite2026.05 | 61.98 | |
| Frozen checkpointModel=Moonlight-16B-A3B-Instruct2026.05 | 59.1 | |
| ST-MoEModel=DeepSeek-V2-Lite2026.05 | 57.34 | |
| Frozen checkpointModel=DeepSeek-V2-Lite2026.05 | 35.42 | |
| Frozen checkpointModel=DeepSeek-MoE-Chat2026.05 | 33.07 |