Question Answering on MuSiQue (OOD)
5.26EMMBT-S
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MBT-SModel=Qwen3 0.6B2026.02 | 5.26 | 3.36 | 2.72 | |
| MBT-RModel=Qwen3 0.6B2026.02 | 4.37 | 2.79 | 2.21 | |
| MBT-SModel=Qwen3 1.7B2026.02 | 2.11 | 1.75 | 1.37 | |
| MBT-SModel=Qwen3 4B2026.02 | 1.83 | 1.5 | 0.85 | |
| MBT-RModel=Qwen3 4B2026.02 | 1.61 | 1.29 | 0.65 | |
| MBT-RModel=Qwen3 1.7B2026.02 | 1.59 | 1.37 | 1.06 | |
| GRPOModel=Qwen3 1.7B2026.02 | 0.88 | 0.7 | 0.56 | |
| GRPOModel=Qwen3 0.6B2026.02 | 0.82 | 0.02 | -0.43 | |
| GRPOModel=Qwen3 4B2026.02 | 0.03 | -0.11 | -0.65 | |
| BaseModel=Qwen3 0.6B2026.02 | 0 | 0 | 0 | |
| BaseModel=Qwen3 1.7B2026.02 | 0 | 0 | 0 | |
| RSModel=Qwen3 1.7B2026.02 | 0 | 0.01 | 0.03 | |
| BaseModel=Qwen3 4B2026.02 | 0 | 0 | 0 | |
| RSModel=Qwen3 0.6B2026.02 | -0.14 | -0.17 | -0.04 | |
| PromptModel=Qwen3 0.6B2026.02 | -0.26 | -0.33 | -0.23 | |
| PromptModel=Qwen3 4B2026.02 | -0.3 | -0.3 | -0.06 | |
| PromptModel=Qwen3 1.7B2026.02 | -0.48 | -0.41 | 0.05 | |
| RSModel=Qwen3 4B2026.02 | -1.28 | -1.3 | -1.36 | |
| gpt-oss-distillModel=Qwen3 4B2026.02 | -1.71 | -2.07 | -2.66 | |
| TokenSkipModel=Qwen3 0.6B2026.02 | -3.82 | -3.69 | -3.7 | |
| LIMOProModel=Qwen3 0.6B2026.02 | -4.26 | -5.08 | -5.26 | |
| gpt-oss-distillModel=Qwen3 1.7B2026.02 | -4.94 | -5.48 | -5.76 | |
| TokenSkipModel=Qwen3 1.7B2026.02 | -5.62 | -5.58 | -5.66 | |
| LIMOProModel=Qwen3 4B2026.02 | -6 | -5.92 | -6.13 | |
| LIMOProModel=Qwen3 1.7B2026.02 | -6.09 | -6.06 | -5.9 | |
| gpt-oss-distillModel=Qwen3 0.6B2026.02 | -6.61 | -8.59 | -9.11 | |
| TokenSkipModel=Qwen3 4B2026.02 | -7.64 | -7.6 | -7.69 |