Long-context reasoning on MRCR ID 4-8K (accuracy)
99.8AccuracyLoRA
Evaluation Results
| Method | Links | |
|---|---|---|
| LoRABase Model=Qwen2.5-7B-Instruct2026.06 | 99.8 | |
| Randomized YaRNBase Model=Qwen2.5-7B-Instruct2026.06 | 96 | |
| Trained YaRNBase Model=Qwen2.5-7B-Instruct2026.06 | 89.1 | |
| RPEBase Model=Qwen2.5-7B-Instruct2026.06 | 81.7 | |
| Trained YaRNBase Model=Olmo3-7B-Instruct2026.06 | 77.9 | |
| Randomized YaRNBase Model=Olmo3-7B-Instruct2026.06 | 77.4 | |
| 0-shotBase Model=Qwen2.5-7B-Instruct2026.06 | 38.9 | |
| 0-shot + YaRNBase Model=Qwen2.5-7B-Instruct2026.06 | 34.6 | |
| 0-shotBase Model=Olmo3-7B-Instruct2026.06 | 21.3 | |
| 0-shot + YaRNBase Model=Olmo3-7B-Instruct2026.06 | 21.3 |