QueryAgent: A Reliable and Efficient Reasoning Framework with Environmental Feedback based Self-Correction
Kernekoncepter
QueryAgent introduces a reliable and efficient reasoning framework for KBQA, outperforming existing few-shot methods by leveraging stepwise self-correction with ERASER.
Resumé
Abstract:
Large Language Models (LLMs) for semantic parsing have shown success but lack reliability and efficiency when hallucinations occur.
QueryAgent addresses these challenges with stepwise self-correction using ERASER, outperforming previous methods on GrailQA and GraphQ.
Introduction:
Recent advances in LLMs for KBQA highlight the need for reliable and efficient reasoning frameworks like QueryAgent.
Existing methods fall short in reliability and efficiency due to error propagation and reliance on black-box models.
Data Extraction:
Experimental results demonstrate QueryAgent's superiority over previous few-shot methods on GrailQA and GraphQ by 7.0 and 15.0 F1 scores.
"Experimental results demonstrate that QueryAgent notably outperforms all previous few-shot methods using only one example on GrailQA and GraphQ by 7.0 and 15.0 F1."