Model Alignment Evaluation on Stoic Alignment Benchmark
32.24Mean ScoreQwen3 Few-shot
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Qwen3 Few-shotBase Model=Qwen3, Evaluation Protocol=Few-shot2026.05 | 32.24 | 32.42 | 2.17 | 100 | 0.22 | 31.81 | — | — | |
| Qwen3 AlphaPO 300Base Model=Qwen3, Alignment Method=AlphaPO, Optimization Steps=3002026.05 | 30.93 | 31 | 2.25 | 100 | 0.23 | 30.49 | 0.001 | — | |
| Qwen3 AlphaPO 200Base Model=Qwen3, Alignment Method=AlphaPO, Optimization Steps=2002026.05 | 29.3 | 29.39 | 2.16 | 100 | 0.22 | 28.88 | 0.001 | — | |
| Qwen3 ORPO 300Base Model=Qwen3, Alignment Method=ORPO, Optimization Steps=3002026.05 | 28.98 | 29.22 | 2.61 | 100 | 0.26 | 28.47 | 0.001 | — | |
| Qwen3 AlphaPO 100Base Model=Qwen3, Alignment Method=AlphaPO, Optimization Steps=1002026.05 | 28.54 | 28.78 | 2.21 | 100 | 0.22 | 28.1 | 0.001 | — | |
| Qwen3 ORPO 100Base Model=Qwen3, Alignment Method=ORPO, Optimization Steps=1002026.05 | 28.46 | 28.56 | 2.01 | 100 | 0.2 | 28.07 | 0.001 | — | |
| Qwen3 ORPO 200Base Model=Qwen3, Alignment Method=ORPO, Optimization Steps=2002026.05 | 28.37 | 28.39 | 2.34 | 100 | 0.23 | 27.92 | 0.001 | — | |
| Qwen3 Zero-shotBase Model=Qwen3, Evaluation Protocol=Zero-shot2026.05 | 27.79 | 28.06 | 2.76 | 100 | 0.28 | 27.25 | 0.001 | — | |
| Llama3 Few-shotBase Model=Llama3, Evaluation Protocol=Few-shot2026.05 | 26.11 | 26.78 | 3.58 | 100 | 0.36 | 25.41 | 0.001 | — | |
| Llama3 ORPO 300Base Model=Llama3, Alignment Method=ORPO, Optimization Steps=3002026.05 | 25.13 | 25.83 | 3.23 | 100 | 0.32 | 24.49 | 0.001 | — | |
| Llama3 AlphaPO 300Base Model=Llama3, Alignment Method=AlphaPO, Optimization Steps=3002026.05 | 21.56 | 22.44 | 4.45 | 100 | 0.45 | 20.69 | 0.001 | — | |
| Llama3 ORPO 200Base Model=Llama3, Alignment Method=ORPO, Optimization Steps=2002026.05 | 21.49 | 21.72 | 4.03 | 100 | 0.4 | 20.7 | 0.001 | — | |
| Llama3 AlphaPO 200Base Model=Llama3, Alignment Method=AlphaPO, Optimization Steps=2002026.05 | 19.82 | 20.39 | 4.34 | 100 | 0.43 | 18.97 | 0.001 | — | |
| Llama3 ORPO 100Base Model=Llama3, Alignment Method=ORPO, Optimization Steps=1002026.05 | 18.66 | 19.17 | 4.52 | 100 | 0.45 | 17.78 | 0.001 | — | |
| Llama3 AlphaPO 100Base Model=Llama3, Alignment Method=AlphaPO, Optimization Steps=1002026.05 | 18.55 | 18.72 | 4.06 | 100 | 0.41 | 17.76 | 0.001 | — | |
| Llama3 Zero-shotBase Model=Llama3, Evaluation Protocol=Zero-shot2026.05 | 17.65 | 17.83 | 4.15 | 100 | 0.42 | 16.83 | 0.001 | — |