VERGE decomposes LLM outputs into atomic claims, autoformalizes them to first-order logic, verifies with SMT solvers and consensus, and refines via minimal correction subsets, yielding 18.7% average uplift on reasoning benchmarks.
(∀x.Guest(x)∧W ine(x) => Cheese(x)) 3.No one who eats cheese likes fish.(∀x.Cheese(x) =>¬LikesF ish(x))
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
VERGE: Formal Refinement and Guidance Engine for Verifiable LLM Reasoning
VERGE decomposes LLM outputs into atomic claims, autoformalizes them to first-order logic, verifies with SMT solvers and consensus, and refines via minimal correction subsets, yielding 18.7% average uplift on reasoning benchmarks.