2605.26457v1 May 26, 2026 cs.SE

Verus-SpecGym: An Agentic Environment for Evaluating Specification Autoformalization

Anmol Agarwal
Anmol Agarwal
Citations: 62
h-index: 3
Pranjal Aggarwal
Pranjal Aggarwal
Citations: 885
h-index: 9
Seungone Kim
Seungone Kim
Citations: 122
h-index: 2
Sean J. Welleck
Sean J. Welleck
Citations: 29
h-index: 4
Natalie Neamtu
Natalie Neamtu
Citations: 6
h-index: 1
Jannis Limperg
Jannis Limperg
Citations: 61
h-index: 2
Cedric Wen Flamant
Cedric Wen Flamant
Amazon
Citations: 47
h-index: 2
Kanna Shimizu
Kanna Shimizu
Citations: 280
h-index: 8
Bryan Parno
Bryan Parno
Citations: 34
h-index: 2

AI coding agents are increasingly used to write real-world software, but ensuring that their outputs are correct remains a fundamental challenge. Formal verification offers a promising path: an agent generates code together with a machine-checked proof, guaranteeing that the code satisfies a formal specification. However, there is no guarantee that the formal spec itself matches the user's intent. In this work, we study specification autoformalization: whether LLM agents can translate informal programming problems into faithful formal specifications. We introduce Verus-SpecBench, a benchmark of 581 spec-writing tasks derived from Codeforces problems targeting Verus, a verifier for Rust, and Verus-SpecGym, an agentic environment in which models interact with Verus, bash, & the filesystem to develop these specs. The central challenge is evaluation: expert-written reference specs are expensive to write, & LLM judges can miss subtle mistakes. We address this by (a) extending Verus's exec_spec mechanism so that generated specs can be executed as Rust code, & (b) testing them against official Codeforces tests & adversarial cases extracted from Codeforces "hacks", which are edge cases written by competitors to break incorrect solutions. On Verus-SpecBench, the strongest model, Gemini 3.1 Pro, solves 77.8% of tasks, other frontier models solve 51.1--57.8% & OSS models reach only 21.5--25.5%. Our analysis of failure modes shows that model-generated specs can omit important input assumptions, accept incorrect outputs, & reject valid ones. We also find that LLM-as-a-judge evaluation misses 26% of the failures our evaluator catches. Overall, our results suggest that spec autoformalization is within reach for frontier agents but remains brittle even on problems where they can already generate correct code. The code, data, & logs can be found at https://github.com/formal-verif-is-cool/verus-spec-gym

0 Citations
0 Influential
27.9657359028 Altmetric
139.8 Score
Original PDF
1

No Analysis Report Yet

This paper hasn't been analyzed by Gemini yet.

Log in to request an AI analysis.

댓글

댓글을 작성하려면 로그인하세요.

아직 댓글이 없습니다. 첫 번째 댓글을 남겨보세요!