Question Answering on TAT-QA (HalRate, Acc)
31.5HalRateVanilla RAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Vanilla RAG2026.04 | 31.5 | 71.4 | |
| FActScore2026.04 | 22.7 | 72.1 | |
| CRAG2026.04 | 20.1 | 73.8 | |
| Self-RAG2026.04 | 18.4 | 74.6 | |
| GPT-4o + CoTPrompting Strategy=Chain of Thought (CoT)2026.04 | 15.2 | 77.1 | |
| FINGROUNDAblation Components=- table ret.2026.04 | 10.6 | 71.8 | |
| FINGROUNDAblation Components=- taxonomy2026.04 | 8.1 | 77.9 | |
| FINGROUNDAblation Components=full2026.04 | 3.8 | 78.3 | |
| FINGROUNDAblation Components=- regen.2026.04 | 3.8 | 72.7 |