Formal Verification on VeriBench
95.2Pass RatePOLCA
Evaluation Results
| Method | Links | |
|---|---|---|
| POLCABackbone=claude-3.5-sonnet, Embedding Model=gemini/text-embedding-004, Budget=50 metric calls per task, Evaluation Protocol=3-step (compilation, unit tests, LLM judge)2026.03 | 95.2 | |
| DSPyBackbone=claude-3.5-sonnet, Budget=50 metric calls per task, Evaluation Protocol=3-step (compilation, unit tests, LLM judge)2026.03 | 88.8 | |
| OpenEvolveBackbone=claude-3.5-sonnet, Budget=50 metric calls per task, Evaluation Protocol=3-step (compilation, unit tests, LLM judge)2026.03 | 73.8 | |
| GEPABackbone=claude-3.5-sonnet, Budget=50 metric calls per task, Evaluation Protocol=3-step (compilation, unit tests, LLM judge)2026.03 | 69.5 |