Class Incremental Learning on OmniBenchmark B-0
82.27Last Accuracy (Alast)Joint fine-tuning
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Joint fine-tuningEvaluation protocol=Joint2026.03 | 82.27 | — | |
| SCL-MGSMIncremental steps (Inc)=102026.03 | 79.9 | 86.92 | |
| SCL-MGSMIncremental steps (Inc)=52026.03 | 79.83 | 86.79 | |
| Joint linear probeEvaluation protocol=Joint2026.03 | 78.56 | — |