Paraphrase Detection on MRPC (Delta Acc and Delta CalLogP)
0Delta AccuracyCalibrated Soft Suffix Learning
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Calibrated Soft Suffix LearningTarget Model=Phi-1.5, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | 0 | -0.1 | |
| UATTarget Model=Phi-1.5, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | 0.24 | |
| AutoPromptTarget Model=Phi-1.5, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | 0.05 | |
| soft promptTarget Model=Phi-1.5, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | 0.17 | |
| UATTarget Model=TinyLlama, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | -0.05 | |
| AutoPromptTarget Model=TinyLlama, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | 0.01 | |
| soft promptTarget Model=TinyLlama, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.01 | 0.02 | |
| AutoPromptTarget Model=Qwen2-1.5B, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.02 | 0.06 | |
| soft promptTarget Model=Qwen2-1.5B, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.02 | -0.11 | |
| soft promptTarget Model=Qwen2-1.5B, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.02 | -0.29 | |
| UATTarget Model=Qwen2-1.5B, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.03 | 1.08 | |
| AutoPromptTarget Model=Qwen2-1.5B, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.03 | -0.68 | |
| UATTarget Model=Phi-1.5, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.38 | |
| AutoPromptTarget Model=Phi-1.5, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.21 | |
| soft promptTarget Model=Phi-1.5, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.3 | |
| UATTarget Model=TinyLlama, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.08 | |
| AutoPromptTarget Model=TinyLlama, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.69 | |
| soft promptTarget Model=TinyLlama, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.04 | 0.66 | |
| Calibrated Soft Suffix LearningTarget Model=Qwen2-1.5B, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.28 | 0.02 | |
| Calibrated Soft Suffix LearningTarget Model=Qwen2-1.5B, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.37 | -0.35 | |
| Calibrated Soft Suffix LearningTarget Model=TinyLlama, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.39 | -0.22 | |
| UATTarget Model=Qwen2-1.5B, Shot Setting=4-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.41 | 1.43 | |
| Calibrated Soft Suffix LearningTarget Model=TinyLlama, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.41 | 0.18 | |
| Calibrated Soft Suffix LearningTarget Model=Phi-1.5, Shot Setting=0-shot, Token Length (K)=4, Seen Model=Qwen2-1.5B-Instruct2025.12 | -0.45 | 0.2 |