arXiv subjects

Yahui Song

Publications and source records attributed to Yahui Song.

4 recordsLinked to original sources

arXiv2025

Agentic Verification of Software Systems

Automatically generated code is gaining traction recently, owing to the prevalence of Large Language Models (LLMs). Further, the AlphaProof initiative has demonstrated the possibility of using AI for general mathematical reasoning. Reasoning about computer programs (software) can be accomplished via general mathematical reasoning; however, it tends to be more structured and richer in contexts. This forms an attractive proposition, since then AI agents can be used to reason about voluminous code that gets generated by AI. In this work, we present a first LLM agent, AutoRocq, for conducting program verification. Unlike past works, which rely on extensive training of LLMs on proof examples, our agent learns on-the-fly and improves the proof via an iterative refinement loop. The iterative improvement of the proof is achieved by the proof agent communicating with the Rocq (formerly Coq) theorem prover to get additional context and feedback. The final result of the iteration is a proof derivation checked by the Rocq theorem prover. In this way, our proof construction involves autonomous collaboration between the proof agent and the theorem prover. This autonomy facilitates the search for proofs and decision-making in deciding on the structure of the proof tree. Experimental evaluation on SV-COMP benchmarks and on Linux kernel modules shows promising efficacy in achieving automated program verification. As automation in code generation becomes more widespread, we posit that our proof agent can be potentially integrated with AI coding agents to achieve a generate and validate loop, thus moving closer to the vision of trusted automatic programming.

cs.SE↗

arXiv2025

Computation Tree Logic Guided Program Repair

Temporal logics like Computation Tree Logic (CTL) have been widely used as expressive formalisms to capture rich behavioral specifications. CTL can express properties such as reachability, termination, invariants and responsiveness, which are difficult to test. This paper suggests a mechanism for the automated repair of infinite-state programs guided by CTL properties. Our produced patches avoid the overfitting issue that occurs in test-suite-guided repair, where the repaired code may not pass tests outside the given test suite. To realize this vision, we propose a repair framework based on Datalog, a widely used domain-specific language for program analysis, which readily supports nested fixed-point semantics of CTL via stratified negation. Specifically, our framework encodes the program and CTL properties into Datalog facts and rules and performs the repair by modifying the facts to pass the analysis rules. Previous research proposed a generic repair mechanism for Datalog-based analysis in the form of Symbolic Execution of Datalog (SEDL). However, SEDL only supports positive Datalog, which is insufficient for expressing CTL properties. Thus, we extended SEDL to make it applicable to stratified Datalog. Moreover, liveness property violations involve infinite computations, which we handle via a novel loop summarization. Our approach achieves analysis accuracy of 56.6% on a small-scale benchmark and 88.5% on a real-world benchmark, outperforming the best baseline performances of 27.7% and 76.9%. Our approach repairs all detected bugs, which is not achieved by existing tools.

cs.SE↗

arXiv2025

Detecting LLM Fact-conflicting Hallucinations Enhanced by Temporal-logic-based Reasoning

Large language models (LLMs) face the challenge of hallucinations -- outputs that seem coherent but are actually incorrect. A particularly damaging type is fact-conflicting hallucination (FCH), where generated content contradicts established facts. Addressing FCH presents three main challenges: 1) Automatically constructing and maintaining large-scale benchmark datasets is difficult and resource-intensive; 2) Generating complex and efficient test cases that the LLM has not been trained on -- especially those involving intricate temporal features -- is challenging, yet crucial for eliciting hallucinations; and 3) Validating the reasoning behind LLM outputs is inherently difficult, particularly with complex logical relationships, as it requires transparency in the model's decision-making process. This paper presents Drowzee, an innovative end-to-end metamorphic testing framework that utilizes temporal logic to identify fact-conflicting hallucinations (FCH) in large language models (LLMs). Drowzee builds a comprehensive factual knowledge base by crawling sources like Wikipedia and uses automated temporal-logic reasoning to convert this knowledge into a large, extensible set of test cases with ground truth answers. LLMs are tested using these cases through template-based prompts, which require them to generate both answers and reasoning steps. To validate the reasoning, we propose two semantic-aware oracles that compare the semantic structure of LLM outputs to the ground truths. Across nine LLMs in nine different knowledge domains, experimental results show that Drowzee effectively identifies rates of non-temporal-related hallucinations ranging from 24.7% to 59.8%, and rates of temporal-related hallucinations ranging from 16.7% to 39.2%.

cs.CL↗

arXiv2023

Staged Specification Logic for Verifying Higher-Order Imperative Programs (Technical Report)

Higher-order functions and imperative states are language features supported by many mainstream languages. Their combination is expressive and useful, but complicates specification and reasoning, due to the use of yet-to-be-instantiated function parameters. One inherent limitation of existing specification mechanisms is its reliance on only two stages: an initial stage to denote the precondition at the start of the method and a final stage to capture the postcondition. Such two-stage specifications force abstract properties to be imposed on unknown function parameters, leading to less precise specifications for higher-order methods. To overcome this limitation, we introduce a novel extension to Hoare logic that supports multiple stages for a call-by-value higher-order language with ML-like local references. Multiple stages allow the behavior of unknown function-type parameters to be captured abstractly as uninterpreted relations; and can also model the repetitive behavior of each recursion as a separate stage. In this paper, we define our staged logic with its semantics, prove its soundness and develop a new automated higher-order verifier, called Heifer, for a core ML-like language.

cs.PL↗