Temporal Reasoning on TemporalBench
Evaluation results
| Method | Links | |
|---|
| MSA-PTTraining Strategy=from-scratch sparse pretraining2026.06 | | 53.4 |
|---|
| MSA-CPTTraining Strategy=sparse continued pretraining2026.06 | | 50.6 |
|---|
| FullTraining Strategy=Full-Attention baseline2026.06 | | 49.4 |
|---|