Reproduce “Specification-Guided Generation for Large Language Models” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Alessandro Previti, Konstantinos Vandikas, Mario Alviano, Mats Carlsson, Albin Larsson Forsberg, Alexandros Nikou, Athanasios Karapantelakis, Aneta Vulgarakis Feljan

SureGen generates structured outputs satisfying supplied logical constraints by filtering each variable’s feasible values before the LLM selects among them. A reasoner excludes choices that cannot extend to a valid assignment, while singleton domains need no decoding call. On conflict-bearing prompts, a judge marked 13/98 backtracking assignments and 92/98 forward assignments non-adherent; this measures judged adherence, not objective user intent.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.