Language Understanding on ARC-C, MMLU, HellaSwag, TruthfulQA (test)
159.215Total ScoreIFD
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| IFDModel=Qwen, SFT Dataset=Alpaca, Retained fraction=0.452026.02 | 159.215 | — | — | — | — | 2.372 | |
| Full SFTModel=Qwen, SFT Dataset=Alpaca, Retained fraction=12026.02 | 156.843 | — | — | — | — | — | |
| TSS++ (E)Model=Qwen, SFT Dataset=NI-20, Retained fraction=0.052026.02 | 155.153 | — | — | — | — | 3.558 | |
| Zero-shotModel=Qwen, Evaluation Protocol=Zero-shot2026.02 | 154.218 | — | — | — | — | — | |
| TSS++ (E)Model=Qwen, SFT Dataset=Dolly, Retained fraction=0.152026.02 | 154.11 | — | — | — | — | 1.716 | |
| Full SFTModel=Qwen, SFT Dataset=Dolly, Retained fraction=12026.02 | 152.394 | — | — | — | — | — | |
| Full SFTModel=Qwen, SFT Dataset=NI-20, Retained fraction=12026.02 | 151.595 | — | — | — | — | — | |
| TSS++ (E)Model=Llama, SFT Dataset=Alpaca, Retained fraction=0.452026.02 | 147.304 | 33.959 | 37.423 | 48.994 | 26.928 | 7.916 | |
| PPLModel=Llama, SFT Dataset=Dolly, Retained fraction=12026.02 | 141.979 | — | — | — | — | 3.788 | |
| RandomModel=Llama, SFT Dataset=Alpaca, Retained fraction=0.452026.02 | 141.241 | 32.594 | 34.831 | 48.357 | 25.459 | — | |
| RandomModel=Llama, SFT Dataset=NI-20, Retained fraction=0.152026.02 | 140.717 | — | — | — | — | 7.805 | |
| Full SFTModel=Llama, SFT Dataset=Alpaca, Retained fraction=12026.02 | 139.388 | — | — | — | — | — | |
| Full SFTModel=Llama, SFT Dataset=Dolly, Retained fraction=12026.02 | 138.191 | — | — | — | — | — | |
| RandomModel=Gemma, SFT Dataset=Alpaca, Retained fraction=0.752026.02 | 136.089 | — | — | — | — | 0.913 | |
| Full SFTModel=Gemma, SFT Dataset=Alpaca, Retained fraction=12026.02 | 135.176 | — | — | — | — | — | |
| Full SFTModel=Llama, SFT Dataset=NI-20, Retained fraction=12026.02 | 132.912 | — | — | — | — | — | |
| RandomModel=Gemma, SFT Dataset=Dolly, Retained fraction=0.752026.02 | 132.526 | — | — | — | — | 1.017 | |
| TSSModel=Gemma, SFT Dataset=NI-20, Retained fraction=0.052026.02 | 132.137 | — | — | — | — | 3.16 | |
| Full SFTModel=Gemma, SFT Dataset=Dolly, Retained fraction=12026.02 | 131.509 | — | — | — | — | — | |
| Zero-shotModel=Gemma, Evaluation Protocol=Zero-shot2026.02 | 129.923 | — | — | — | — | — | |
| Full SFTModel=Gemma, SFT Dataset=NI-20, Retained fraction=12026.02 | 128.977 | — | — | — | — | — | |
| Zero-shotModel=Llama, Evaluation Protocol=Zero-shot2026.02 | 127.493 | — | — | — | — | — |