LLM Steering on Persona
81.75Average Trait ScoreUNISTEER
Evaluation Results
| Method | Links | |
|---|---|---|
| UNISTEERTarget LLM=Qwen2.5-7B2026.05 | 81.75 | |
| CAATarget LLM=Qwen2.5-7B2026.05 | 80.67 | |
| UNISTEERTarget LLM=Qwen2.5-1.5B2026.05 | 77.67 | |
| UNISTEERTarget LLM=Llama-3.2-1B2026.05 | 72.03 | |
| CAATarget LLM=Llama-3.2-1B2026.05 | 67.33 | |
| LoReFTTarget LLM=Qwen2.5-1.5B2026.05 | 54.67 | |
| ODESteerTarget LLM=Qwen2.5-1.5B2026.05 | 47.67 | |
| LoReFTTarget LLM=Qwen2.5-7B2026.05 | 42.78 | |
| RepETarget LLM=Qwen2.5-7B2026.05 | 36.67 | |
| RepETarget LLM=Llama-3.2-1B2026.05 | 33.67 | |
| RepETarget LLM=Qwen2.5-1.5B2026.05 | 31.1 | |
| CAATarget LLM=Qwen2.5-1.5B2026.05 | 28 | |
| LoReFTTarget LLM=Llama-3.2-1B2026.05 | 20.76 | |
| OriginalTarget LLM=Llama-3.2-1B2026.05 | 19.37 | |
| ODESteerTarget LLM=Llama-3.2-1B2026.05 | 16.6 | |
| ODESteerTarget LLM=Qwen2.5-7B2026.05 | 11.67 | |
| OriginalTarget LLM=Qwen2.5-7B2026.05 | 5.1 | |
| OriginalTarget LLM=Qwen2.5-1.5B2026.05 | 4.1 |