Syntactic Correctness on C++
100Syntactic CorrectnessDream
Evaluation Results
| Method | Links | |
|---|---|---|
| DreamSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamCoderSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamCoderSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamCoderSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamCoderSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DreamCoderSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=16, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=64, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=128, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| LLaDASteps=256, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DiffuCoderSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DiffuCoderSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DiffuCoderSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 100 | |
| DiffuCoderSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 100 | |
| DreamSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.8 | |
| DreamCoderSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.8 | |
| DreamCoderSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.8 | |
| LLaDASteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.8 | |
| DiffuCoderSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.8 | |
| DreamSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.6 | |
| DreamSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.6 | |
| LLaDASteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.6 | |
| DreamCoderSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.4 | |
| DreamCoderSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.4 | |
| LLaDASteps=32, Decoding Strategy=EPIC, Witness-based Recovery=true2026.05 | 99.4 | |
| LLaDASteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 99.4 | |
| LLaDASteps=256, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 99.4 | |
| DreamCoderSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.2 | |
| LLaDASteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=true2026.05 | 99.2 | |
| LLaDASteps=256, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 96.3 | |
| DreamSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 95.1 | |
| DreamSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 95.1 | |
| DreamSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 94.5 | |
| DreamSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 94.5 | |
| DiffuCoderSteps=256, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 93.9 | |
| DiffuCoderSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 93.7 | |
| LLaDASteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 92.7 | |
| LLaDASteps=128, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 92.7 | |
| DiffuCoderSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 91.3 | |
| DreamSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 90.9 | |
| DreamSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 90.2 | |
| DreamSteps=128, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 89 | |
| DreamCoderSteps=256, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 86 | |
| DreamCoderSteps=256, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 85 | |
| DreamCoderSteps=256, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 84.1 | |
| DreamSteps=256, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 82.9 | |
| LLaDASteps=128, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 82.9 | |
| DreamCoderSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 81.1 | |
| DreamSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 80.5 | |
| DreamSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 80.5 | |
| DreamSteps=64, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 80.5 | |
| DreamCoderSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 80.5 | |
| DreamCoderSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 78.7 | |
| DreamCoderSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 77.4 | |
| DiffuCoderSteps=128, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 74.4 | |
| DreamCoderSteps=128, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 73.8 | |
| DreamCoderSteps=64, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 72.6 | |
| DiffuCoderSteps=128, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 72.6 | |
| DreamSteps=32, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 67.7 | |
| DiffuCoderSteps=64, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 67.7 | |
| LLaDASteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 67.1 | |
| LLaDASteps=64, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 67.1 | |
| DiffuCoderSteps=64, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 67.1 | |
| DreamCoderSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 66.5 | |
| DreamCoderSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 65.9 | |
| DiffuCoderSteps=128, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 60.4 | |
| DreamSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 54.3 | |
| DreamCoderSteps=32, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 54.3 | |
| DiffuCoderSteps=64, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 53.7 | |
| DreamSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 53 | |
| DiffuCoderSteps=32, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 51.8 | |
| DiffuCoderSteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 50 | |
| LLaDASteps=64, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 42.7 | |
| DiffuCoderSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 40.9 | |
| DreamCoderSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 39 | |
| DiffuCoderSteps=16, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 39 | |
| DiffuCoderSteps=32, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 36 | |
| DreamSteps=16, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 35.4 | |
| LLaDASteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 34.1 | |
| DreamCoderSteps=16, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 33.5 | |
| LLaDASteps=16, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 32.9 | |
| LLaDASteps=32, Decoding Strategy=CFG-constrained baseline, Witness-based Recovery=false2026.05 | 31.7 | |
| LLaDASteps=32, Decoding Strategy=EPIC, Witness-based Recovery=false2026.05 | 31.1 | |
| DiffuCoderSteps=16, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 25 | |
| DreamCoderSteps=16, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 17.7 | |
| LLaDASteps=32, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 9.1 | |
| LLaDASteps=16, Decoding Strategy=Unconstrained decoding, Witness-based Recovery=false2026.05 | 5.5 |