An LLM-plus-symbolic-validator loop cuts structural requirement errors to 0.39% under a small model (0% under a frontier model) and quantifies ~25% of LLM choices as valid but indeterminate.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.SE 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Model-Driven Requirements Configuration with Three-Valued Uncertainty Scoring
An LLM-plus-symbolic-validator loop cuts structural requirement errors to 0.39% under a small model (0% under a frontier model) and quantifies ~25% of LLM choices as valid but indeterminate.