Generation of and Debugging with Logical Pre and Postconditions

大型语言模型能否将自然语言意图转化为形式化方法后置条件？

Can Large Language Models Transform Natural Language Intent into Formal Method Postconditions?

本文提出 nl2postcond 任务，用 LLM 将自然语言意图转为形式化后置条件，并通过正确性、判别力指标及 Defects4J 实验验证其有效性。

Constraint-Based Synthesis of Coupling Proofs

Proof by coupling is a classical technique for proving properties about pairs of randomized algorithms by carefully relating (or coupling) two probabilistic executions. In this paper, we show how to automatically construct such proofs for probabilistic programs. First, we present f-coupled postconditions, an abstraction describing two correlated program executions. Second, we show how properties of f-coupled postconditions can imply various probabilistic properties of the original programs. Third, we demonstrate how to reduce the proof-search problem to a purely logical synthesis problem of th…

JavaTA: A Logic-based Debugger for Java

This paper presents a logic based approach to debugging Java programs. In contrast with traditional debugging we propose a debugging methodology for Java programs using logical queries on individual execution states and also over the history of execution. These queries were arrived at by a systematic study of errors in object-oriented programs in our earlier research. We represent the salient events during the execution of a Java program by a logic database, and implement the queries as logic programs. Such an approach allows us to answer a number of useful and interesting queries about a Java…

循环不变式生成：推理优化LLM与SMT求解器的混合框架

Loop Invariant Generation: A Hybrid Framework of Reasoning optimised LLMs and SMT Solvers

本研究提出将推理优化大语言模型与Z3 SMT求解器结合的生成-检验框架，利用反例迭代精化循环不变式，在Code2Inv基准的133个任务上实现100%覆盖率，显著优于此前最优结果，证明LLM的逻辑推理能力可助力循环不变式自动综合。