Coding on CRUXEval-O
87.5ScoreLLaDA2.1-flash
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LLaDA2.1-flashInference Mode=Q Mode2026.02 | 87.5 | 380 | |
| Qwen3-30B-A3B-Inst-25072026.02 | 86.75 | 100 | |
| LLaDA2.1-flashInference Mode=S Mode2026.02 | 85.25 | 654 | |
| LLaDA2.0-flash2026.02 | 85.12 | 321 | |
| Ling-flash-2.02026.02 | 82.75 | 100 | |
| Ling-mini-2.02026.02 | 76.12 | — | |
| Qwen3-8Bno think=true2026.02 | 74.06 | — | |
| LLaDA2.1-minimode=Q Mode2026.02 | 73.75 | 3.35 | |
| LLaDA2.0-mini2026.02 | 71.62 | 2.78 | |
| LLaDA2.1-minimode=S Mode2026.02 | 70.62 | 5.85 | |
| DSRMidtraining Schedule=DSR2026.05 | 40.88 | — | |
| Cosine-decayMidtraining Schedule=Cos2026.05 | 40.38 | — | |
| WSDMidtraining Schedule=WSD2026.05 | 40 | — | |
| Unified 15B MoEMidtraining Corpus=1k→8k2026.05 | 39.75 | — | |
| Unified 15B MoEMidtraining Corpus=4k→8k2026.05 | 37.5 | — | |
| Unified 15B MoEMidtraining Corpus=4k2026.05 | 35.62 | — | |
| Unified 15B MoEPretraining Corpus=4k→8k2026.05 | 28.62 | — | |
| Unified 15B MoEPretraining Corpus=4k2026.05 | 28.12 | — | |
| Unified 15B MoEPretraining Corpus=1k→8k2026.05 | 27.25 | — |