paper-with-me

Papers

EquiBench: Benchmarking Large Language Models' Understanding of Program Semantics via Equivalence Checking

2025-02-18 · Anjiang Wei, Jiannan Cao, Ran Li, Hongyu Chen, Yuhui Zhang, Ziheng Wang, YuAn Liu, Thiago S. F. X. Teixeira, Diyi Yang, Ke Wang, Alex Aiken

As large language models (LLMs) become integral to code-related tasks, a central question emerges: do LLMs truly understand program execution semantics? We introduce EquiBench, a new benchmark for evaluating LLMs through equivalence checking, i.e., determining whether two programs produce identical outputs for all possible inputs. Unlike prior code generation benchmarks, this task directly tests a model's understanding of code execution semantics. EquiBench consists of 2400 program pairs across four languages and six categories. These pairs are generated through program analysis, compiler scheduling, and superoptimization, ensuring high-confidence labels, nontrivial difficulty, and full automation. The transformations span syntactic edits, structural modifications, and algorithmic changes, covering a broad spectrum of semantic variation. We evaluate 19 state-of-the-art LLMs and find that in the most challenging categories, the best accuracies are 63.8% and 76.2%, only modestly above the 50% random baseline. Further analysis reveals that models often rely on syntactic similarity rather than exhibiting robust reasoning over execution semantics, highlighting fundamental limitations.

📄 PDF Abstract BibTeX arXiv:2502.12466

Code (0)

등록된 구현이 없습니다.

Tasks

BenchmarkingBinary ClassificationCode GenerationScheduling

Similar Papers 제목 키워드 기반

Benchmarking Language Models for Code Syntax Understanding

2022-10-26 · Da Shen, Xinyun Chen, Chenguang Wang, Koushik Sen 외

Pre-trained language models have demonstrated impressive performance in both natural language processing and program understanding, which represent the input as a token sequence without explicitly modeling its structure.…

Benchmarking

ThrowBench: Benchmarking LLMs by Predicting Runtime Exceptions

2025-03-06 · Julian Aron Prenner, Romain Robbes

Modern Large Language Models (LLMs) have shown astounding capabilities of code understanding and synthesis. In order to assess such capabilities, several benchmarks have been devised (e.g., HumanEval). However, most benc…

BenchmarkingHumanEval

Improving LLM Code Reasoning via Semantic Equivalence Self-Play with Formal Verification

2026-04-18 · Antonio Valerio Miceli Barone, Poon Tsz Nok arxiv

We introduce a self-play framework for semantic equivalence in Haskell, utilizing formal verification to guide adversarial training between a generator and an evaluator. The framework leverages Liquid Haskell proofs for …

BLAST: Benchmarking LLMs with ASP-based Structured Testing

2026-04-24 · Manuel Alejandro Borroto Santana, Erica Coppolillo, Francesco Calimeri, Giuseppe Manco 외 arxiv

Large Language Models (LLMs) have demonstrated remarkable performance across a broad spectrum of tasks, including natural language understanding, dialogue systems, and code generation. Despite evident progress, less atte…

Natural Language UnderstandingCode Generation

DependEval: Benchmarking LLMs for Repository Dependency Understanding

2025-03-09 · Junjia Du, Yadi Liu, Hongcheng Guo, Jiawei Wang 외

While large language models (LLMs) have shown considerable promise in code generation, real-world software development demands advanced repository-level reasoning. This includes understanding dependencies, project struct…

BenchmarkingCode Generation