Prompt-based Value Alignment on Schwartz Value Prompts (test)
0.83GainGemini
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GeminiSteer Value=Hedonism, Shots=8s2026.02 | 0.83 | 0.13 | 0.07 | 0.01 | |
| GeminiSteer Value=Hedonism, Shots=2s2026.02 | 0.76 | 0.12 | 0.08 | 0.02 | |
| GeminiSteer Value=Hedonism, Shots=4s2026.02 | 0.65 | 0.13 | 0.13 | 0.02 | |
| QwenSteer Value=Stimulation, Shots=8s2026.02 | 0.56 | 0.12 | 0.15 | 0.02 | |
| DeepSeekSteer Value=Stimulation, Shots=4s2026.02 | 0.55 | 0.12 | 0.16 | 0.02 | |
| QwenSteer Value=Stimulation, Shots=2s2026.02 | 0.52 | 0.12 | 0.14 | 0.02 | |
| DeepSeekSteer Value=Stimulation, Shots=8s2026.02 | 0.51 | 0.11 | 0.13 | 0.02 | |
| QwenSteer Value=Stimulation, Shots=4s2026.02 | 0.5 | 0.12 | 0.13 | 0.01 | |
| QwenSteer Value=Hedonism, Shots=8s2026.02 | 0.45 | 0.09 | 0.1 | 0.01 | |
| DeepSeekSteer Value=Hedonism, Shots=8s2026.02 | 0.44 | 0.1 | 0.13 | 0.01 | |
| DeepSeekSteer Value=Hedonism, Shots=4s2026.02 | 0.42 | 0.1 | 0.13 | 0.01 | |
| QwenSteer Value=Hedonism, Shots=2s2026.02 | 0.41 | 0.1 | 0.12 | 0.01 | |
| GeminiSteer Value=Stimulation, Shots=4s2026.02 | 0.4 | 0.13 | 0.11 | 0.02 | |
| QwenSteer Value=Hedonism, Shots=4s2026.02 | 0.39 | 0.1 | 0.12 | 0.02 | |
| GeminiSteer Value=Stimulation, Shots=8s2026.02 | 0.38 | 0.11 | 0.12 | 0.02 | |
| DeepSeekSteer Value=Stimulation, Shots=2s2026.02 | 0.35 | 0.1 | 0.09 | 0.02 | |
| DeepSeekSteer Value=Hedonism, Shots=2s2026.02 | 0.34 | 0.11 | 0.08 | 0.01 | |
| GeminiSteer Value=Stimulation, Shots=2s2026.02 | 0.32 | 0.12 | 0.1 | 0.02 | |
| GeminiSteer Value=Power, Shots=4s2026.02 | 0.26 | 0.13 | 0.14 | 0.02 | |
| GeminiSteer Value=Security, Shots=4s2026.02 | 0.24 | 0.13 | 0.09 | 0.02 | |
| GeminiSteer Value=Power, Shots=8s2026.02 | 0.23 | 0.12 | 0.13 | 0.02 | |
| GeminiSteer Value=Power, Shots=2s2026.02 | 0.22 | 0.12 | 0.1 | 0.02 | |
| GeminiSteer Value=Security, Shots=2s2026.02 | 0.2 | 0.13 | 0.12 | 0.01 | |
| GeminiSteer Value=Security, Shots=8s2026.02 | 0.2 | 0.12 | 0.11 | 0.02 | |
| QwenSteer Value=Security, Shots=8s2026.02 | 0.16 | 0.11 | 0.13 | 0.02 | |
| GPTSteer Value=Security, Shots=4s2026.02 | 0.14 | 0.13 | 0.11 | 0.02 | |
| QwenSteer Value=Security, Shots=2s2026.02 | 0.14 | 0.11 | 0.13 | 0.01 | |
| GPTSteer Value=Power, Shots=2s2026.02 | 0.12 | 0.14 | 0.09 | 0.02 | |
| QwenSteer Value=Security, Shots=4s2026.02 | 0.12 | 0.11 | 0.15 | 0.01 | |
| DeepSeekSteer Value=Power, Shots=4s2026.02 | 0.11 | 0.09 | 0.08 | 0.01 | |
| DeepSeekSteer Value=Power, Shots=8s2026.02 | 0.11 | 0.11 | 0.07 | 0.02 | |
| GPTSteer Value=Security, Shots=8s2026.02 | 0.09 | 0.14 | 0.12 | 0.01 | |
| GPTSteer Value=Power, Shots=4s2026.02 | 0.09 | 0.13 | 0.08 | 0.02 | |
| QwenSteer Value=Power, Shots=4s2026.02 | 0.09 | 0.11 | 0.1 | 0.02 | |
| QwenSteer Value=Power, Shots=8s2026.02 | 0.09 | 0.1 | 0.09 | 0.01 | |
| DeepSeekSteer Value=Security, Shots=2s2026.02 | 0.08 | 0.11 | 0.15 | 0.01 | |
| GPTSteer Value=Power, Shots=8s2026.02 | 0.08 | 0.13 | 0.11 | 0.02 | |
| GPTSteer Value=Security, Shots=2s2026.02 | 0.07 | 0.15 | 0.12 | 0.02 | |
| QwenSteer Value=Power, Shots=2s2026.02 | 0.07 | 0.11 | 0.11 | 0.02 | |
| DeepSeekSteer Value=Power, Shots=2s2026.02 | 0.06 | 0.12 | 0.1 | 0.02 | |
| GPTSteer Value=Hedonism, Shots=8s2026.02 | 0.06 | 0.13 | 0.14 | 0.02 | |
| DeepSeekSteer Value=Security, Shots=8s2026.02 | 0.05 | 0.11 | 0.14 | 0.01 | |
| GPTSteer Value=Hedonism, Shots=2s2026.02 | 0.05 | 0.14 | 0.15 | 0.02 | |
| DeepSeekSteer Value=Security, Shots=4s2026.02 | 0.02 | 0.1 | 0.14 | 0.02 | |
| GPTSteer Value=Hedonism, Shots=4s2026.02 | 0.02 | 0.14 | 0.16 | 0.01 | |
| GPTSteer Value=Stimulation, Shots=2s2026.02 | -0.11 | 0.14 | 0.14 | 0.02 | |
| GPTSteer Value=Stimulation, Shots=8s2026.02 | -0.14 | 0.12 | 0.15 | 0.02 | |
| GPTSteer Value=Stimulation, Shots=4s2026.02 | -0.18 | 0.13 | 0.13 | 0.02 |