Misaligned Task Learning on Legal In-domain
0.87MisalignmentPersona Vectors
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Persona VectorsModel=Qwen2.5-32B2025.08 | 0.87 | 2.07 | |
| KLModel=Qwen2.5-32B2025.08 | 7.67 | 2.67 | |
| Interleaving++Model=Qwen2.5-32B2025.08 | 27.17 | 25.8 | |
| Interleaving+Model=Qwen2.5-32B2025.08 | 28.03 | 24.27 | |
| InterleavingModel=Qwen2.5-32B2025.08 | 28.37 | 25.2 | |
| MisalignedModel=Qwen2.5-32B2025.08 | 31.33 | 23.67 |