PICARD: Parsing Incrementally for Constrained Auto-Regressive Decoding from Language Models

Benchmark Model Rank Results
text-to-sql-on-spider-1T5-3B+PICARD#1Exact Match Accuracy (in Dev): 75.5Execution Accuracy (in Dev): 79.3
text-to-sql-on-spider-1T5-3B#4Exact Match Accuracy (in Dev): 71.5Execution Accuracy (in Dev): 74.4
dialogue-state-tracking-on-cosqlT5-3B + PICARD#2question match accuracy: 54.6interaction match accuracy: 23.7
semantic-parsing-on-spiderT5-3B + PICARD#5Accuracy: 71.9
text-to-sql-on-spiderT5-3B + PICARD#11Execution Accuracy (Test): 75.1Exact Match Accuracy (Test): 71.9…