Soundness Verification on Certificate Soundness (test)
100Soundness RateNeural Hybrid Router
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Neural Hybrid RouterVerdict=NO, Certificate Kind=bounded_proof, Verification Protocol=cycle-consistency verification2026.06 | 100 | 35 | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=bounded_proof, Verification Protocol=cycle-consistency verification2026.06 | 100 | 35 | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=bounded_proof, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 100 | — | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=k-induction, Verification Protocol=cycle-consistency verification2026.06 | 100 | 35 | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=inductive_invariant, Verification Protocol=cycle-consistency verification2026.06 | 97 | 34 | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=lasso, Verification Protocol=cycle-consistency verification2026.06 | 97 | 34 | |
| Neural Hybrid RouterVerdict=NO, Certificate Kind=lasso, Verification Protocol=cycle-consistency verification2026.06 | 93 | 33 | |
| Neural Hybrid RouterVerdict=NO, Certificate Kind=witness_pair, Verification Protocol=cycle-consistency verification2026.06 | 90 | 32 | |
| Neural Hybrid RouterVerdict=All, Certificate Kind=Total, Verification Protocol=cycle-consistency verification2026.06 | 90 | 378 | |
| Neural Hybrid RouterVerdict=NO, Certificate Kind=inductive_invariant, Verification Protocol=cycle-consistency verification2026.06 | 87 | 30 | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=inductive_invariant, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 87 | — | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=k-induction, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 87 | — | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=witness_pair, Verification Protocol=cycle-consistency verification2026.06 | 87 | 29 | |
| Neural Hybrid RouterVerdict=YES, Certificate Kind=reachability, Verification Protocol=cycle-consistency verification2026.06 | 83 | 29 | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=lasso, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 80 | — | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=reachability, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 80 | — | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=lasso, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 77 | — | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=witness_pair, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 77 | — | |
| Opus 4.5 / Opus 4.5Verdict=YES, Certificate Kind=witness_pair, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 77 | — | |
| Opus 4.5 / Opus 4.5Verdict=All, Certificate Kind=Total, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 76.1 | — | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=inductive_invariant, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 73 | — | |
| Neural Hybrid RouterVerdict=NO, Certificate Kind=k-induction, Verification Protocol=cycle-consistency verification2026.06 | 73 | 26 | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=k-induction, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 73 | — | |
| Neural Hybrid RouterVerdict=NO, Certificate Kind=reachability, Verification Protocol=cycle-consistency verification2026.06 | 73 | 26 | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=reachability, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 60 | — | |
| Opus 4.5 / Opus 4.5Verdict=NO, Certificate Kind=bounded_proof, Verification Protocol=cycle-consistency verification, Prompting Strategy=few-shot2026.06 | 43 | — |