Regulatory Question Answering on Regulatory-domain QA (test)
52.8EMCOMPLIANCENLP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| COMPLIANCENLPRetrieval=With retrieval2026.04 | 52.8 | 71.9 | |
| COMPLIANCENLP w/o MiniCheckRetrieval=With retrieval2026.04 | 52.1 | 71 | |
| COMPLIANCENLP w/o multi-taskRetrieval=With retrieval2026.04 | 50.3 | 69.1 | |
| LLaMA-3-70B*Retrieval=With retrieval2026.04 | 49.1 | 67.4 | |
| GPT-4o + RAGRetrieval=With retrieval2026.04 | 48.1 | 66.8 | |
| COMPLIANCENLP w/o KG rerankingRetrieval=With retrieval2026.04 | 47.6 | 66.2 | |
| LLaMA-3-8B + RAGRetrieval=With retrieval2026.04 | 47.4 | 65.9 | |
| GPT-4 + RAGRetrieval=With retrieval2026.04 | 46.3 | 64.7 | |
| GPT-4oRetrieval=Without retrieval, Prompting=5-shot2026.04 | 43.7 | 61.3 | |
| GPT-4Retrieval=Without retrieval, Prompting=5-shot2026.04 | 41.2 | 58.6 | |
| RIRAGRetrieval=With retrieval2026.04 | 38.9 | 54.2 |