Unverified60% confidenceBenchmarkExact time
NL2SQL technology has achieved over 80% accuracy on benchmarks such as Spider, but accuracy drops significantly on complex operations such as multi-table JOINs and subqueries.
2
Sources
60%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
UnverifiedGPT-4等大模型在Spider等基准测试上的零样本或少样本SQL生成准确率已超过80%76% similarUnverified在Spider基准上，早期单模型方案准确率约为70%，GPT-4等大模型将其提升到80%以上，多智能体协作策略的系统已能达到85%甚至更高的执行准确率72% similarUnverifiedOn the Spider benchmark, the accuracy of state-of-the-art models rose from under 20% in 2018 to over 80%72% similarUnverifiedOn the Mixed-WM38 dataset (containing 6,866 wafer maps), the multi-region prompting strategy achieved 92.9% macro-average accuracy and 93.8% overall accuracy.65% similarUnverifiedIn experiments on 4.2 GB of waveforms, 18450 signals and 150 expert-annotated queries, the BTTF research team achieved 95.33% execution accuracy, with Multi-Agent quality scores 17.3% higher than single Agent, average latency increasing to about 1.96 times, and static SVA check accuracy of 84.62%.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/691020API
curl https://kongchang.com/api/v1/knowledge/claims/691020MCP
get_claim(id=691020)