Multilingual Table Question Answering on MultiTEND German
25.65EM (Exact Match)MultiLink
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MultiLink2025.02 | 25.65 | 64.73 | 73.73 | 58.9 | 85.27 | 73.55 | |
| SMART2025.02 | 18.3 | 59.35 | 63.86 | 44.22 | 83.72 | 73.68 | |
| Fine-tuned LlamaBackbone=Llama, Evaluation Protocol=Fine-tuned2025.02 | 15.78 | 56.14 | 62.35 | 41.44 | 81.44 | 70.72 | |
| RAG for LLMEvaluation Protocol=RAG2025.02 | 14.33 | 60.17 | 67.47 | 48.59 | 74.31 | 70.41 | |
| Few-shot LLMEvaluation Protocol=Few-shot2025.02 | 10.65 | 53.83 | 58.38 | 37.08 | 65.31 | 66.35 | |
| Zero-shot LLMEvaluation Protocol=Zero-shot2025.02 | 0.54 | 50.43 | 60.07 | 34.91 | 57.11 | 60.14 |