Formal Verification on VeruSAGE-Bench
62.5IronKV AccuracySTRACE
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| STRACEPrompt Setting=STRACE (Ours)2026.07 | 62.5 | 88.9 | 100 | 31.7 | 61.5 | 58.5 | |
| Summary-based SelectionPrompt Setting=Summary-based Selection2026.07 | 58.3 | 61.1 | 70 | 29.3 | 30.8 | 45.3 | |
| GEPAPrompt Setting=GEPA2026.07 | 54.2 | 61.1 | 80 | 29.3 | 46.2 | 47.2 | |
| Base AgentPrompt Setting=Base Agent2026.07 | 41.6 | 66.7 | 80 | 20 | 46.2 | 42.5 | |
| Naive Few-shotPrompt Setting=Naive Few-shot2026.07 | 37.5 | 61.1 | 80 | 24.4 | 30.8 | 39.6 | |
| Failure-Aware RAGPrompt Setting=Failure-Aware RAG2026.07 | 37.5 | 66.7 | 70 | 29.3 | 38.5 | 42.5 | |
| Retrieval-based SelectionPrompt Setting=Retrieval-based Selection2026.07 | 37.5 | 66.7 | 90 | 22 | 38.5 | 41.5 |