Beyond Solver Verdicts: Generative Reward Models for Autoformalization
Generative Verification (GenV) is introduced, which distills an offline Z3-equivalence oracle into a reference-free, continuous reference-equivalence score by repurposing the language model's native vocabulary space and theoretically proves that structural, verdict-only verification heuristics are mathematically bounde...
Vikash Singh, Debargha Ganguly, Aman Goel et al.
· 0 citations