Distribution Alignment on Sentiment Even
0.075MAEKL-Optimized Fine-Tuning
Evaluation Results
| Method | Links | |
|---|---|---|
| KL-Optimized Fine-TuningModel=Qwen2.5-1.5B-Instruct2026.04 | 0.075 | |
| IFTModel=Qwen2.5-1.5B-Instruct2026.04 | 0.099 | |
| KL-Optimized Fine-TuningModel=Qwen2.5-7B-Instruct2026.04 | 0.114 | |
| Zero-shotModel=Llama-3.1-8B-Instruct2026.04 | 0.159 | |
| IFTModel=Qwen2.5-7B-Instruct2026.04 | 0.162 | |
| KL-Optimized Fine-TuningModel=Llama-3.2-1B-Instruct2026.04 | 0.182 | |
| KL-Optimized Fine-TuningModel=Llama-3.1-8B-Instruct2026.04 | 0.199 | |
| IFTModel=Llama-3.2-1B-Instruct2026.04 | 0.231 | |
| IFTModel=Llama-3.1-8B-Instruct2026.04 | 0.236 | |
| Zero-shotModel=Llama-3.2-1B-Instruct2026.04 | 0.269 | |
| DPOModel=Qwen2.5-1.5B-Instruct2026.04 | 0.27 | |
| Zero-shotModel=Qwen2.5-1.5B-Instruct2026.04 | 0.287 | |
| DPOModel=Llama-3.2-1B-Instruct2026.04 | 0.314 | |
| PE-ExplicitModel=Llama-3.2-1B-Instruct2026.04 | 0.341 | |
| DPOModel=Qwen2.5-7B-Instruct2026.04 | 0.364 | |
| DPOModel=Llama-3.1-8B-Instruct2026.04 | 0.394 | |
| Zero-shotModel=Qwen2.5-7B-Instruct2026.04 | 0.405 | |
| PE-ExplicitModel=Llama-3.1-8B-Instruct2026.04 | 0.435 | |
| PE-ExplicitModel=Qwen2.5-1.5B-Instruct2026.04 | 0.443 | |
| PE-ExplicitModel=Qwen2.5-7B-Instruct2026.04 | 0.444 |