Question Answering on SciQA (test)
80.6AccuracyDLR
Evaluation Results
| Method | Links | |
|---|---|---|
| DLRModel=Qwen3-8B2026.05 | 80.6 | |
| DLRModel=Qwen3-4B2026.05 | 72.3 | |
| TokenAssortedModel=Qwen3-8B2026.05 | 69.7 | |
| TokenAssortedModel=Qwen3-4B2026.05 | 68 | |
| DLR (C=1)Model=Qwen3-8B2026.05 | 66.4 | |
| PauseTokenModel=Qwen3-8B2026.05 | 65.2 | |
| DLRModel=Qwen3-1.7B2026.05 | 64.1 | |
| DLR (C=1)Model=Qwen3-4B2026.05 | 64.1 | |
| PauseTokenModel=Qwen3-4B2026.05 | 63.8 | |
| DLRModel=Llama3.2-3B2026.05 | 63.3 | |
| SFTModel=Qwen3-8B2026.05 | 61.8 | |
| DLR (C=1)Model=Qwen3-1.7B2026.05 | 60 | |
| SFTModel=Qwen3-4B2026.05 | 59.4 | |
| DLR (C=1)Model=Llama3.2-3B2026.05 | 59.2 | |
| PauseTokenModel=Llama3.2-3B2026.05 | 57.6 | |
| SFTModel=Llama3.2-3B2026.05 | 56.7 | |
| SFTModel=Qwen3-1.7B2026.05 | 56.4 | |
| PauseTokenModel=Qwen3-1.7B2026.05 | 56.2 | |
| DLRModel=Qwen3-0.6B2026.05 | 55.3 | |
| DLRModel=Llama3.2-1B2026.05 | 49.1 | |
| DLR (C=1)Model=Qwen3-0.6B2026.05 | 48.9 | |
| SFTModel=Qwen3-0.6B2026.05 | 48 | |
| PauseTokenModel=Qwen3-0.6B2026.05 | 47.9 | |
| DLR (C=1)Model=Llama3.2-1B2026.05 | 43.7 | |
| SFTModel=Llama3.2-1B2026.05 | 38 | |
| TokenAssortedModel=Llama3.2-3B2026.05 | 34.6 | |
| PauseTokenModel=Llama3.2-1B2026.05 | 34 | |
| TokenAssortedModel=Llama3.2-1B2026.05 | 33.6 | |
| TokenAssortedModel=Qwen3-1.7B2026.05 | 17.3 | |
| TokenAssortedModel=Qwen3-0.6B2026.05 | 13.1 |