Table Question Answering on SST-QA
86.1AccuracyOracle
Evaluation results
| Method | Links | |
|---|---|---|
| OracleDescription=Theoretical upper bound via ideal selection between Textual and Symbolic reasoning2026.04 | 86.1 | |
| ASTRA (Adaptive Selection)Reasoning mode=Adaptive2026.04 | 81.9 | |
| ASTRA (Textual Reasoning)Reasoning mode=Textual2026.04 | 79.8 | |
| o3Category=Foundation Models2026.04 | 78.2 | |
| EEDPCategory=Prompting/Tool-augmented Methods2026.04 | 76.8 | |
| ASTRA (Symbolic Reasoning)Reasoning mode=Symbolic2026.04 | 75.3 | |
| GraphOTTERCategory=Intermediate-Representation Methods2026.04 | 71.5 | |
| ST-RaptorCategory=Intermediate-Representation Methods2026.04 | 71.1 | |
| E5Category=Prompting/Tool-augmented Methods2026.04 | 70.2 | |
| GPT-4oCategory=Foundation Models2026.04 | 66.4 | |
| DeepSeek-V3Category=Foundation Models2026.04 | 63.2 | |
| TableLlamaCategory=Table-Specific Adapted Models2026.04 | 40.4 |