Prompt Steering
91Sum-1Olmo-3.1-32B +CD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Olmo-3.1-32B +CDBackbone=Olmo-3.1-32B, Method=CD, alpha=22026.01 | 91 | 97 | |
| Olmo-3.1-32B +CDBackbone=Olmo-3.1-32B, Method=CD, alpha=12026.01 | 88 | 96 | |
| Llama-3.1-8B +CDBackbone=Llama-3.1-8B, Method=CD, alpha=22026.01 | 87 | 95 | |
| Qwen-2.5-32B +CDBackbone=Qwen-2.5-32B, Method=CD, alpha=22026.01 | 83 | 95 | |
| Olmo-3.1-32B +CDBackbone=Olmo-3.1-32B, Method=CD, alpha=02026.01 | 83 | 93 | |
| Llama-3.1-8B +CDBackbone=Llama-3.1-8B, Method=CD, alpha=12026.01 | 82 | 93 | |
| Olmo-3.1-32B +PromptingBackbone=Olmo-3.1-32B, Method=Prompting2026.01 | 82 | 93 | |
| Qwen-2.5-14B +CDBackbone=Qwen-2.5-14B, Method=CD, alpha=22026.01 | 80 | 94 | |
| Qwen-2.5-32B +CDBackbone=Qwen-2.5-32B, Method=CD, alpha=12026.01 | 80 | 94 | |
| Qwen-2.5-14B +CDBackbone=Qwen-2.5-14B, Method=CD, alpha=12026.01 | 76 | 94 | |
| Qwen-2.5-32B +PromptingBackbone=Qwen-2.5-32B, Method=Prompting2026.01 | 76 | 91 | |
| Qwen-2.5-32B +CDBackbone=Qwen-2.5-32B, Method=CD, alpha=02026.01 | 76 | 92 | |
| Llama-3.1-8B +PromptingBackbone=Llama-3.1-8B, Method=Prompting2026.01 | 76 | 91 | |
| Llama-3.1-8B +CDBackbone=Llama-3.1-8B, Method=CD, alpha=02026.01 | 76 | 91 | |
| Qwen-2.5-7B +CDBackbone=Qwen-2.5-7B, Method=CD, alpha=22026.01 | 73 | 91 | |
| Qwen-2.5-7B +CDBackbone=Qwen-2.5-7B, Method=CD, alpha=12026.01 | 71 | 90 | |
| Qwen-2.5-14B +PromptingBackbone=Qwen-2.5-14B, Method=Prompting2026.01 | 69 | 92 | |
| Qwen-2.5-14B +CDBackbone=Qwen-2.5-14B, Method=CD, alpha=02026.01 | 69 | 92 | |
| Qwen-2.5-7B +PromptingBackbone=Qwen-2.5-7B, Method=Prompting2026.01 | 60 | 85 | |
| Qwen-2.5-7B +CDBackbone=Qwen-2.5-7B, Method=CD, alpha=02026.01 | 60 | 85 |