AuraTracer智迹闻
中文

EVENT DOSSIER

AxQM: A Textbook-Scale Benchmark for Formal Proof Synthesis in a Library of Finite-Dimensional Quantum Mechanics

2026-09-07 12:00 Science 🔥 42.2 heat score
1sources
1days unfolding
42.2heat score
2mentions
SummaryAI generated

AxQM benchmark was released on September 7, 2026, aiming to evaluate the formal proofing capabilities. The benchmark includes 1,019 verifiable tasks from the textbook “Quantum Computing and Quantum Information,” with a scale approximately four times larger than existing levels. All tasks are derived from the complete formalization of the official parts of the textbook and are built based on a custom finite-dimensional quantum mechanics Lean library. The solutions remain private; scoring is performed by a deterministic check within the Lean kernel, ensuring that the proofs pass, there are no sorry statements, and no new axioms are introduced.

Related eventsRELATED EVENTS
Key entitiesKEY ENTITIES
AxQMLean

Coverage · reports per dayLANGUAGE SPLIT

Entity relations
AxQM × Lean1

SignalsSIGNALS

Keyword heat
  • AxQM1
  • Lean1

All reports (1)SOURCES

A arXiv cs.AI en 2026-09-07 12:00

AxQM: A Textbook-Scale Benchmark for Formal Proof Synthesis in a Library of Finite-Dimensional Quantum Mechanics

AxQM 发布,包含来自《量子计算与量子信息》教材的 1019 个可验证证明合成任务。该基准基于自定义的有限维量子力学 Lean 库构建,是物理学中按任务数量最大的证明合成基准,规模约为现有水平的四倍。所有任务均源自对教材正式部分的近完整形式化,且解决方案保持私有。基准评分由 Lean 内核执行确定性检查,确保证明编译通过、不包含 sorry 声明且不引入新公理。