Code Generation on MBPP+ strict evaluation
0.468AccuracyTuneShift-KD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TuneShift-KDSource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Target LoRA2026.03 | 0.468 | 0.061 | |
| TuneShift-KDSource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Target (no LoRA)2026.03 | 0.407 | — | |
| Trans-LoRASource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Target LoRA2026.03 | 0.406 | 0.085 | |
| Trans-LoRASource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Target LoRA2026.03 | 0.344 | 0.027 | |
| TuneShift-KDSource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Target LoRA2026.03 | 0.339 | 0.023 | |
| Trans-LoRASource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Source LoRA2026.03 | 0.339 | — | |
| TuneShift-KDSource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Source LoRA2026.03 | 0.324 | — | |
| Trans-LoRASource Model=Gemma-2B, Target Model=Gemma-7B, Evaluation Setup=Target (no LoRA)2026.03 | 0.321 | — | |
| Trans-LoRASource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Target (no LoRA)2026.03 | 0.317 | — | |
| TuneShift-KDSource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Target (no LoRA)2026.03 | 0.316 | — | |
| TuneShift-KDSource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Source LoRA2026.03 | 0.257 | — | |
| Trans-LoRASource Model=Llama2-7B, Target Model=Llama2-13B, Evaluation Setup=Source LoRA2026.03 | 0.25 | — |