Text-to-NoSQL on MultiTEND average of 6 languages (test)
25.54Exact Match (EM)MultiLink
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MultiLink2025.02 | 25.54 | 64.01 | 73.17 | 59.12 | 85.66 | 74.01 | |
| SMART2025.02 | 17.23 | 59.12 | 61.94 | 43.36 | 85.05 | 76.37 | |
| Fine-tuned Llamalearning_strategy=fine-tuned, backbone=Llama2025.02 | 15.64 | 55.9 | 58.44 | 40.95 | 81.41 | 70.46 | |
| RAG for LLMretrieval=RAG2025.02 | 14.32 | 59.79 | 67.19 | 47.95 | 74.22 | 69.8 | |
| Few-shot LLMprompting_strategy=few-shot2025.02 | 10.82 | 56.17 | 61.12 | 36.69 | 64.43 | 64.71 | |
| Zero-shot LLMprompting_strategy=zero-shot2025.02 | 0.48 | 49.35 | 59.36 | 32.44 | 55.98 | 59.22 |