Value Alignment on GLOBALOPINIONQA VITAL Distributional 6 values
0.18JS DistanceVISPA
Evaluation results
| Method | Links | |
|---|---|---|
| VISPABase Model=LLaMA2-7B, Steered Backbone=LLaMA2-7B2026.01 | 0.18 | |
| VISPABase Model=LLaMA2-7B, Steered Backbone=LLaMA3-8B2026.01 | 0.195 | |
| VISPABase Model=LLaMA3-8B, Steered Backbone=LLaMA3-8B2026.01 | 0.21 | |
| VISPABase Model=LLaMA3-8B, Steered Backbone=LLaMA2-7B2026.01 | 0.216 | |
| VISPABase Model=Gemma-7B, Steered Backbone=LLaMA3-8B2026.01 | 0.228 | |
| EthosBase Model=ChatGPT2026.01 | 0.231 | |
| VISPABase Model=Gemma-7B, Steered Backbone=LLaMA2-7B2026.01 | 0.238 | |
| EthosBase Model=Qwen2.5-7B2026.01 | 0.253 | |
| EthosBase Model=LLaMA3-8B2026.01 | 0.254 | |
| VISPABase Model=LLaMA2-13B, Steered Backbone=LLaMA2-7B2026.01 | 0.256 | |
| EthosBase Model=LLaMA2-13B2026.01 | 0.259 | |
| EthosBase Model=LLaMA2-7B2026.01 | 0.261 | |
| VISPABase Model=ChatGPT, Steered Backbone=LLaMA3-8B2026.01 | 0.271 | |
| ModPluralBase Model=ChatGPT2026.01 | 0.274 | |
| VISPABase Model=LLaMA2-13B, Steered Backbone=LLaMA3-8B2026.01 | 0.277 | |
| EthosBase Model=Qwen2.5-14B2026.01 | 0.278 | |
| ModPluralBase Model=LLaMA3-8B2026.01 | 0.281 | |
| VISPABase Model=Qwen2.5-7B, Steered Backbone=LLaMA3-8B2026.01 | 0.297 | |
| VISPABase Model=Qwen2.5-14B, Steered Backbone=LLaMA3-8B2026.01 | 0.298 | |
| ModPluralBase Model=LLaMA2-13B2026.01 | 0.305 | |
| EthosBase Model=Gemma-7B2026.01 | 0.307 | |
| VanillaBase Model=LLaMA2-13B2026.01 | 0.312 | |
| ModPluralBase Model=Qwen2.5-14B2026.01 | 0.312 | |
| VISPABase Model=ChatGPT, Steered Backbone=LLaMA2-7B2026.01 | 0.318 | |
| VISPABase Model=Qwen2.5-14B, Steered Backbone=LLaMA2-7B2026.01 | 0.319 | |
| VISPABase Model=Qwen2.5-7B, Steered Backbone=LLaMA2-7B2026.01 | 0.32 | |
| ModPluralBase Model=Qwen2.5-7B2026.01 | 0.329 | |
| VanillaBase Model=LLaMA3-8B2026.01 | 0.329 | |
| ModPluralBase Model=Gemma-7B2026.01 | 0.333 | |
| VanillaBase Model=LLaMA2-7B2026.01 | 0.35 | |
| VanillaBase Model=Qwen2.5-14B2026.01 | 0.366 | |
| VanillaBase Model=ChatGPT2026.01 | 0.374 | |
| ModPluralBase Model=LLaMA2-7B2026.01 | 0.395 | |
| MoEBase Model=LLaMA3-8B2026.01 | 0.399 | |
| MoEBase Model=LLaMA2-13B2026.01 | 0.405 | |
| VanillaBase Model=Gemma-7B2026.01 | 0.408 | |
| MoEBase Model=LLaMA2-7B2026.01 | 0.439 | |
| VanillaBase Model=Qwen2.5-7B2026.01 | 0.441 | |
| MoEBase Model=ChatGPT2026.01 | 0.441 | |
| MoEBase Model=Qwen2.5-14B2026.01 | 0.486 | |
| MoEBase Model=Qwen2.5-7B2026.01 | 0.504 | |
| MoEBase Model=Gemma-7B2026.01 | 0.52 |