Modules are building, update coming soon! Time-Semantic Architecture
MMLU-Pro Benchmark Results
- Model ID: Vi-LAB/KenT_0.5B_Base
- Benchmark: TIGER-Lab/MMLU-Pro (Source Data)
- Overall Accuracy: 11.66% (1,403 / 12,032 correct)
| Category | Accuracy (%) | Correct / Total |
|---|---|---|
| philosophy | 14.43% | 72 / 499 |
| law | 13.81% | 152 / 1101 |
| history | 13.39% | 51 / 381 |
| economics | 13.03% | 110 / 844 |
| health | 12.35% | 101 / 818 |
| physics | 11.78% | 153 / 1299 |
| other | 11.58% | 107 / 924 |
| chemistry | 11.31% | 128 / 1132 |
| engineering | 11.25% | 109 / 969 |
| math | 10.73% | 145 / 1351 |
| biology | 10.60% | 76 / 717 |
| computer science | 10.49% | 43 / 410 |
| business | 10.39% | 82 / 789 |
| psychology | 9.27% | 74 / 798 |