Academic Journal
ReproCopilot: LLM-Driven Failure Reproduction with Dynamic Refinement
| Title: | ReproCopilot: LLM-Driven Failure Reproduction with Dynamic Refinement |
|---|---|
| Authors: | Tanakorn Leesatapornwongsa, Fazle Faisal, Suman Nath |
| Source: | Proceedings of the ACM on Software Engineering. 2:2920-2943 |
| Publisher Information: | Association for Computing Machinery (ACM), 2025. |
| Publication Year: | 2025 |
| Description: | Failure reproduction is a crucial step for debugging software systems, but it is often challenging and time-consuming, especially when the failures are caused by complex inputs, states, or environments. In this paper, we present ReproCopilot, a tool that leverages program analysis and a large language model (LLM) to generate a workload (i.e., code and inputs) that can reproduce a given failure. ReproCopilot proposes two novel techniques: state-oriented code generation and dynamic refinement. These techniques can iteratively guide the LLM with program analysis feedback until the generated workload can successfully reproduce the target failure. We evaluate ReproCopilot on 50 real-world failures from 17 open-source projects, and show that it can reproduce 76% of them, significantly outperforming the-state-of-the-art solutions. |
| Document Type: | Article |
| Language: | English |
| ISSN: | 2994-970X |
| DOI: | 10.1145/3729399 |
| Accession Number: | edsair.doi...........b16d8466d64794e7f6ef4d0a09c565db |
| Database: | OpenAIRE |
| FullText | Text: Availability: 0 |
|---|---|
| Header | DbId: edsair DbLabel: OpenAIRE An: edsair.doi...........b16d8466d64794e7f6ef4d0a09c565db RelevancyScore: 1033 AccessLevel: 3 PubType: Academic Journal PubTypeId: academicJournal PreciseRelevancyScore: 1032.51025390625 |
| IllustrationInfo | |
| Items | – Name: Title Label: Title Group: Ti Data: ReproCopilot: LLM-Driven Failure Reproduction with Dynamic Refinement – Name: Author Label: Authors Group: Au Data: <searchLink fieldCode="AR" term="%22Tanakorn+Leesatapornwongsa%22">Tanakorn Leesatapornwongsa</searchLink><br /><searchLink fieldCode="AR" term="%22Fazle+Faisal%22">Fazle Faisal</searchLink><br /><searchLink fieldCode="AR" term="%22Suman+Nath%22">Suman Nath</searchLink> – Name: TitleSource Label: Source Group: Src Data: <i>Proceedings of the ACM on Software Engineering</i>. 2:2920-2943 – Name: Publisher Label: Publisher Information Group: PubInfo Data: Association for Computing Machinery (ACM), 2025. – Name: DatePubCY Label: Publication Year Group: Date Data: 2025 – Name: Abstract Label: Description Group: Ab Data: Failure reproduction is a crucial step for debugging software systems, but it is often challenging and time-consuming, especially when the failures are caused by complex inputs, states, or environments. In this paper, we present ReproCopilot, a tool that leverages program analysis and a large language model (LLM) to generate a workload (i.e., code and inputs) that can reproduce a given failure. ReproCopilot proposes two novel techniques: state-oriented code generation and dynamic refinement. These techniques can iteratively guide the LLM with program analysis feedback until the generated workload can successfully reproduce the target failure. We evaluate ReproCopilot on 50 real-world failures from 17 open-source projects, and show that it can reproduce 76% of them, significantly outperforming the-state-of-the-art solutions. – Name: TypeDocument Label: Document Type Group: TypDoc Data: Article – Name: Language Label: Language Group: Lang Data: English – Name: ISSN Label: ISSN Group: ISSN Data: 2994-970X – Name: DOI Label: DOI Group: ID Data: 10.1145/3729399 – Name: AN Label: Accession Number Group: ID Data: edsair.doi...........b16d8466d64794e7f6ef4d0a09c565db |
| PLink | https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsair&AN=edsair.doi...........b16d8466d64794e7f6ef4d0a09c565db |
| RecordInfo | BibRecord: BibEntity: Identifiers: – Type: doi Value: 10.1145/3729399 Languages: – Text: English PhysicalDescription: Pagination: PageCount: 24 StartPage: 2920 Titles: – TitleFull: ReproCopilot: LLM-Driven Failure Reproduction with Dynamic Refinement Type: main BibRelationships: HasContributorRelationships: – PersonEntity: Name: NameFull: Tanakorn Leesatapornwongsa – PersonEntity: Name: NameFull: Fazle Faisal – PersonEntity: Name: NameFull: Suman Nath IsPartOfRelationships: – BibEntity: Dates: – D: 19 M: 06 Type: published Y: 2025 Identifiers: – Type: issn-print Value: 2994970X – Type: issn-locals Value: edsair Numbering: – Type: volume Value: 2 Titles: – TitleFull: Proceedings of the ACM on Software Engineering Type: main |
| ResultId | 1 |