PExA: Parallel Exploration Agent Achieves 70.2% on Spider 2.0
Researchers have introduced PExA (Parallel Exploration Agent), a new LLM-based framework for text-to-SQL generation that reformulates the task through software test coverage. Instead of a sequential process, PExA prepares the original query with a suite of test cases using simpler, atomic SQLs executed in parallel to ensure semantic coverage. The final SQL is generated only after sufficient information is gathered from the explored test case SQLs. Validated on the Spider 2.0 benchmark, PExA achieves a new state-of-the-art execution accuracy of 70.2%.
Key facts
- PExA stands for Parallel Exploration Agent.
- It is designed for complex text-to-SQL tasks.
- The framework uses parallel execution of atomic SQLs as test cases.
- It addresses the latency-performance trade-off in LLM-based agents.
- The final SQL is generated after iterative test case coverage.
- PExA achieved 70.2% execution accuracy on Spider 2.0.
- Spider 2.0 is a state-of-the-art benchmark for text-to-SQL.
- The research is published on arXiv under Computer Science > Artificial Intelligence.
Entities
Institutions
- arXiv