Question Answering on MMLU (Answer-letter accuracy)
74Answer-letter AccuracySPD
Evaluation results
| Method | Links | |
|---|---|---|
| SPDBackbone=Qwen2.5-14B-Instruct2026.05 | 74 | |
| SSDBackbone=Qwen2.5-14B-Instruct2026.05 | 73.5 | |
| Base modelBackbone=Qwen2.5-14B-Instruct2026.05 | 72.5 | |
| PSRBackbone=Qwen2.5-14B-Instruct2026.05 | 72.5 | |
| PSRBackbone=Qwen2.5-7B-Instruct2026.05 | 72 | |
| SPDBackbone=Qwen2.5-7B-Instruct2026.05 | 68.5 | |
| SSDBackbone=Qwen2.5-7B-Instruct2026.05 | 67.5 | |
| SPDBackbone=Qwen3-4B-Instruct2026.05 | 65.5 | |
| Base modelBackbone=Qwen2.5-7B-Instruct2026.05 | 65 | |
| SSDBackbone=Qwen3-4B-Instruct2026.05 | 64 | |
| Base modelBackbone=Qwen3-4B-Instruct2026.05 | 63 | |
| SPDBackbone=Llama-3.1-8B-Instruct2026.05 | 62.5 | |
| PSRBackbone=Llama-3.1-8B-Instruct2026.05 | 62 | |
| PSRBackbone=Qwen3-4B-Instruct2026.05 | 61 | |
| SSDBackbone=Llama-3.1-8B-Instruct2026.05 | 60.5 | |
| Base modelBackbone=Llama-3.1-8B-Instruct2026.05 | 60 | |
| SPDBackbone=Qwen2.5-0.5B-Instruct2026.05 | 49 | |
| SSDBackbone=Qwen2.5-0.5B-Instruct2026.05 | 48 | |
| Base modelBackbone=Qwen2.5-0.5B-Instruct2026.05 | 46 | |
| PSRBackbone=Qwen2.5-0.5B-Instruct2026.05 | 43 |