Text-to-SQL Parsing on Ambrosia Unambiguous (test)
88.7RecallIntentRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| IntentRLModel=4B Instruct, Output Format=Interpretation-Answer Pairs2025.11 | 88.7 | 74.7 | |
| SFTModel=4B Instruct, Output Format=Answer-only2025.11 | 78 | 73.4 | |
| SFTModel=4B Instruct, Output Format=Interpretation-Answer Pairs2025.11 | 77.7 | 74.7 | |
| IntentPromptModel=235B MoE Instruct, Output Format=Interpretation-Answer Pairs2025.11 | 74.1 | 57.8 | |
| IntentPromptModel=235B MoE Instruct, Output Format=Answer-only2025.11 | 69.3 | 64.1 | |
| IntentPromptModel=235B MoE Thinking, Output Format=Answer-only2025.11 | 61.8 | 48.4 | |
| IntentPromptModel=4B Instruct, Output Format=Answer-only2025.11 | 58.3 | 57.2 | |
| IntentPromptModel=4B Thinking, Output Format=Answer-only2025.11 | 58.3 | 56.1 | |
| IntentPromptModel=235B MoE Thinking, Output Format=Interpretation-Answer Pairs2025.11 | 53.7 | 35.9 | |
| IntentPromptModel=4B Thinking, Output Format=Interpretation-Answer Pairs2025.11 | 25.2 | 19.7 | |
| IntentPromptModel=4B Instruct, Output Format=Interpretation-Answer Pairs2025.11 | 22 | 18.8 |