Common sense Reasoning on mCSQA (test)
90.8EN AccuracyL8b_T
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| L8b_TMerging Technique=TIES, Density (d)=0.52026.01 | 90.8 | 83.2 | 76 | 77.4 | 72.4 | 80 | |
| L8b_DTSMerging Technique=DARE-TIES-KnOTS, Density (d)=12026.01 | 89.8 | 82.4 | 75.6 | 75.8 | 73.6 | 79.4 | |
| L8b_TSMerging Technique=TIES-KnOTS, Density (d)=12026.01 | 89.8 | 82.4 | 75.6 | 75.8 | 73.6 | 79.4 | |
| L8b_COMBProtocol=COMB, Backbone=Llama-3.1-8b-Instruct2026.01 | 89.6 | 84 | 75.4 | 75.4 | 73.2 | 79.5 | |
| L8b_INDVProtocol=INDV, Backbone=Llama-3.1-8b-Instruct2026.01 | 87.6 | 83.6 | 77 | 75.8 | 72.4 | 79.3 | |
| L8b_DTMerging Technique=DARE-TIES, Density (d)=12026.01 | 87.4 | 82.2 | 75.2 | 77.6 | 70.8 | 78.6 |