arXiv subjects

Yihe Li

Publications and source records attributed to Yihe Li.

3 recordsLinked to original sources

arXiv2026

When is LLM-Based Program Reasoning Correct? A Completion Semantics for LLM-Based Code Inference

Due to token and cognitive limits, Large Language Models (LLMs) typically perform program reasoning over incomplete code fragments/prompts rather than complete programs. Such reasoning therefore must rely on {assumptions about omitted code and context. As a result, the meaning of an inference over a program fragment is not absolute, but depends on an implicit completion model describing how the fragment may be refined into a complete program. In this paper, we introduce completion semantics for LLM-based program reasoning. We formalize incomplete programs as denoting a space of possible refinements and define the correctness of existential inferences relative to a completion model. Under this view, a reported bug is correct whenever there exists a completion within the model that witnesses the bug. This perspective explains why many LLM-generated reports are neither simply correct nor incorrect, but instead depend on assumptions about omitted context. We have instantiated our approach in the form of a witness-generation workflow that concretizes completions underlying an inference by constructing executable refinements of the original program fragment. Witnesses serve both as evidence for existential claims and as a mechanism for exposing the assumptions required to support them. We evaluate our approach on real-world LLM-generated bug reports and program-analysis tasks. Our results show that witness generation effectively distinguishes inferences supported by plausible completions from those requiring unrealistic assumptions, providing a practical mechanism for validating reasoning over incomplete programs.

cs.PL↗

arXiv2026

Persistent Iterators with Value Semantics

Iterators are a fundamental programming abstraction for traversing and modifying elements in containers in mainstream imperative languages such as C++. Iterators provide a uniform access mechanism that hides low-level implementation details of the underlying data structure. However, iterators over mutable containers suffer from well-known hazards including invalidation, aliasing, data races, and subtle side effects. Immutable data structures, as used in functional programming languages, avoid the pitfalls of mutation but rely on a very different programming model based on recursion and higher-order combinators rather than iteration. However, these combinators are not always well-suited to expressing certain algorithms, and recursion can expose implementation details of the underlying data structure. In this paper, we propose persistent iterators -- a new abstraction that reconciles the familiar iterator-based programming style of imperative languages with the semantics of persistent data structures. A persistent iterator snapshots the version of its underlying container at creation, ensuring safety against invalidation and aliasing. Iterator operations operate on the iterator-local copy of the container, giving true value semantics: variables can be rebound to new persistent values while previous versions remain accessible. We implement our approach in the form of LibFPP -- a C++ container library providing persistent vectors, maps, sets, strings, and other abstractions as persistent counterparts to the Standard Template Library (STL). Our evaluation shows that LibFPP retains the expressiveness of iterator-based programming, eliminates iterator-invalidation, and achieves asymptotic complexities comparable to STL implementations. Our design targets use cases where persistence and safety are desired, while allowing developers to retain familiar iterator-based programming patterns.

cs.PL↗

arXiv2025

Large Language Model Powered Symbolic Execution

Large Language Models (LLMs) have emerged as a promising alternative to traditional static program analysis methods, such as symbolic execution, offering the ability to reason over code directly without relying on theorem provers or SMT solvers. However, LLMs are also inherently approximate by nature, and therefore face significant challenges in relation to the accuracy and scale of analysis in real-world applications. Such issues often necessitate the use of larger LLMs with higher token limits, but this requires enterprise-grade hardware (GPUs) and thus limits accessibility for many users. In this paper, we propose LLM-based symbolic execution -- a novel approach that enhances LLM inference via a path-based decomposition of the program analysis tasks into smaller (more tractable) subtasks. The core idea is to generalize path constraints using a generic code-based representation that the LLM can directly reason over, and without translation into another (less-expressive) formal language. We implement our approach in the form of AutoBug, an LLM-based symbolic execution engine that is lightweight and language-agnostic, making it a practical tool for analyzing code that is challenging for traditional approaches. We show that AutoBug can improve both the accuracy and scale of LLM-based program analysis, especially for smaller LLMs that can run on consumer-grade hardware.

cs.PL↗