Long-Context Understanding on Ruler 4K
100Subtask 1 Score (s1)LLaDA-8B-Instruct
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LLaDA-8B-InstructContext Length=4K2026.05 | 100 | 100 | 100 | 92.4 | 100 | 99.8 | 96.92 | 30.56 | 79.18 | 78.8 | — | — | |
| LLaDA-8B-Instruct + Fast-dLLM & WaveFilterContext Length=4K, Optimization=Fast-dLLM & WaveFilter2026.05 | 99.8 | 100 | 100 | 85.8 | 98.2 | 98.6 | 94 | 1.64 | 77.35 | 76.8 | — | — | |
| LLaDA-8B-Instruct + Fast-dLLMContext Length=4K, Optimization=Fast-dLLM2026.05 | 99.6 | 100 | 100 | 88 | 97.55 | 99.55 | 93.08 | 1.58 | 77.32 | 77.4 | — | — | |
| LLaDA-8B-Instruct + Elastic-Cache & WaveFilterContext Length=4K, Optimization=Elastic-Cache & WaveFilter2026.05 | 99.6 | 100 | 100 | 87.4 | 97.15 | 99.78 | 96.78 | 16.95 | 80.75 | 77.8 | — | — | |
| LLaDA-8B-Instruct + Elastic-CacheContext Length=4K, Optimization=Elastic-Cache2026.05 | 99.2 | 100 | 100 | 90.2 | 96.85 | 99.75 | 96.48 | 16.5 | 80.37 | 77.2 | — | — |