Sentiment analysis on GOE (test)
85.16AccuracyDPUA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DPUABackbone=LLaMa-3.1-Instruct (8B)2026.05 | 85.16 | 85.14 | 0.1303 | 0.7854 | |
| w/ GRPOBackbone=LLaMa-3.1-Instruct (8B)2026.05 | 84.59 | 84.59 | 0.211 | 0.6903 | |
| DPUABackbone=Qwen-3-Instruct (8B)2026.05 | 83.78 | 83.78 | 0.1466 | 0.7697 | |
| w/ SFTBackbone=LLaMa-3.1-Instruct (8B)2026.05 | 83.62 | 83.62 | 0.1878 | 0.7265 | |
| w/ GRPO*Backbone=LLaMa-3.1-Instruct (8B)2026.05 | 83.43 | 83.43 | 0.1534 | 0.7175 | |
| w/ GRPOBackbone=Qwen-3-Instruct (8B)2026.05 | 83.23 | 83.23 | 0.2474 | 0.6851 | |
| w/ GRPO*Backbone=Qwen-3-Instruct (8B)2026.05 | 83.14 | 83.13 | 0.1622 | 0.7136 | |
| w/ SFTBackbone=Qwen-3-Instruct (8B)2026.05 | 82.63 | 82.62 | 0.2458 | 0.7107 | |
| w/ SFT*Backbone=Qwen-3-Instruct (8B)2026.05 | 82.43 | 82.41 | 0.1558 | 0.7355 | |
| w/ SFT*Backbone=LLaMa-3.1-Instruct (8B)2026.05 | 81.7 | 81.69 | 0.136 | 0.7671 | |
| few-shotBackbone=Qwen-3-Instruct (8B)2026.05 | 79.01 | 78.7 | 0.2357 | 0.6494 | |
| few-shotBackbone=LLaMa-3.1-Instruct (8B)2026.05 | 77.19 | 77.09 | 0.2326 | 0.604 | |
| zero-shotBackbone=Qwen-3-Instruct (8B)2026.05 | 76.94 | 76.09 | 0.235 | 0.6571 | |
| zero-shotBackbone=LLaMa-3.1-Instruct (8B)2026.05 | 75.68 | 75.33 | 0.221 | 0.6002 |