Towards Understanding the Characteristics of Code Generation Errors Made by Large Language Models

Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation errors that LLMs can produce. To bridge the gap, we conducted an in-depth analysis of code generation errors across six representative LLMs...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Proceedings / International Conference on Software Engineering S. 2587 - 2599
Hauptverfasser: Wang, Zhijie, Zhou, Zijie, Song, Da, Huang, Yuheng, Chen, Shengmai, Ma, Lei, Zhang, Tianyi
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 26.04.2025
Schlagworte:
ISSN:1558-1225
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Abstract Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation errors that LLMs can produce. To bridge the gap, we conducted an in-depth analysis of code generation errors across six representative LLMs on the HumanEval dataset. Specifically, we first employed open coding and thematic analysis to distill a comprehensive taxonomy of code generation errors. We analyzed two dimensions of error characteristics-semantic characteristics and syntactic characteristics. Our analysis revealed that LLMs often made non-trivial, multi-line code generation errors in various locations and with various root causes. We further analyzed the correlation between these errors and task complexity as well as test pass rate. Our findings highlighted several challenges in locating and fixing code generation errors made by LLMs. In the end, we discussed several future directions to address these challenges.
AbstractList Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation errors that LLMs can produce. To bridge the gap, we conducted an in-depth analysis of code generation errors across six representative LLMs on the HumanEval dataset. Specifically, we first employed open coding and thematic analysis to distill a comprehensive taxonomy of code generation errors. We analyzed two dimensions of error characteristics-semantic characteristics and syntactic characteristics. Our analysis revealed that LLMs often made non-trivial, multi-line code generation errors in various locations and with various root causes. We further analyzed the correlation between these errors and task complexity as well as test pass rate. Our findings highlighted several challenges in locating and fixing code generation errors made by LLMs. In the end, we discussed several future directions to address these challenges.
Author Zhang, Tianyi
Huang, Yuheng
Ma, Lei
Wang, Zhijie
Zhou, Zijie
Song, Da
Chen, Shengmai
Author_xml – sequence: 1
  givenname: Zhijie
  surname: Wang
  fullname: Wang, Zhijie
  email: zhijie.wang@ualberta.ca
  organization: University of Alberta,Edmonton,AB,Canada
– sequence: 2
  givenname: Zijie
  surname: Zhou
  fullname: Zhou, Zijie
  email: zijiez4@illinois.edu
  organization: University of Illinois Urbana-Champaign,Champaign,IL,USA
– sequence: 3
  givenname: Da
  surname: Song
  fullname: Song, Da
  email: dsong4@ualberta.ca
  organization: University of Alberta,Edmonton,AB,Canada
– sequence: 4
  givenname: Yuheng
  surname: Huang
  fullname: Huang, Yuheng
  email: yuhenghuang42@g.ecc.u-tokyo.ac.jp
  organization: The University of Tokyo,Tokyo,Japan
– sequence: 5
  givenname: Shengmai
  surname: Chen
  fullname: Chen, Shengmai
  email: chen3301@purdue.edu
  organization: Purdue University,West Lafayette,IN,USA
– sequence: 6
  givenname: Lei
  surname: Ma
  fullname: Ma, Lei
  email: ma.lei@acm.org
  organization: The University of Tokyo,Tokyo,Japan
– sequence: 7
  givenname: Tianyi
  surname: Zhang
  fullname: Zhang, Tianyi
  email: tianyi@purdue.edu
  organization: Purdue University,West Lafayette,IN,USA
BookMark eNotkEFPAjEUhKvRRED-AYf-gcX3ttvSHs0GkQTiQTiTx_YV1mDXtGsM_95N9DIz-TKZw4zFXewiCzFDmCOCe1rX70utVbWYl1DqOQBauBFTt3BWKdSgjcNbMUKtbYFlqR_EOOcPADCVcyNBu-6Hks9yHz2n3FP0bTzJ_syyPlOipufU5r5tsuyCrDvPcsWRE_VtF-UypS5luaUBH69yQ-nEg8bTNw1hO7Qv-VHcB7pknv77ROxflrv6tdi8rdb186ag0kBfENqjQSRCUxmrQ8WWVaOUBU_oAwMo03BgbcBbWNhA1dFRcF5DUwIpNRGzv92WmQ9fqf2kdD0MH5XOaVS_2tdYDw
CODEN IEEPAD
ContentType Conference Proceeding
DBID 6IE
6IH
CBEJK
RIE
RIO
DOI 10.1109/ICSE55347.2025.00180
DatabaseName IEEE Electronic Library (IEL) Conference Proceedings
IEEE Proceedings Order Plan (POP) 1998-present by volume
IEEE Xplore All Conference Proceedings
IEEE Electronic Library (IEL)
IEEE Proceedings Order Plans (POP) 1998-present
DatabaseTitleList
Database_xml – sequence: 1
  dbid: RIE
  name: IEEE Electronic Library (IEL)
  url: https://ieeexplore.ieee.org/
  sourceTypes: Publisher
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
EISBN 9798331505691
EISSN 1558-1225
EndPage 2599
ExternalDocumentID 11029951
Genre orig-research
GroupedDBID -~X
.4S
.DC
29O
5VS
6IE
6IF
6IH
6IK
6IL
6IM
6IN
8US
AAJGR
AAWTH
ABLEC
ADZIZ
ALMA_UNASSIGNED_HOLDINGS
ARCSS
AVWKF
BEFXN
BFFAM
BGNUA
BKEBE
BPEOZ
CBEJK
CHZPO
EDO
FEDTE
I-F
IEGSK
IJVOP
IPLJI
M43
OCL
RIE
RIL
RIO
ID FETCH-LOGICAL-a260t-a18b611aa164685f4e8e3c3380da1dfe0036cefe560d8078fa4b9af9d50c20a33
IEDL.DBID RIE
ISICitedReferencesCount 0
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001538318100202&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
IngestDate Wed Aug 27 01:40:13 EDT 2025
IsPeerReviewed false
IsScholarly true
Language English
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-a260t-a18b611aa164685f4e8e3c3380da1dfe0036cefe560d8078fa4b9af9d50c20a33
PageCount 13
ParticipantIDs ieee_primary_11029951
PublicationCentury 2000
PublicationDate 2025-April-26
PublicationDateYYYYMMDD 2025-04-26
PublicationDate_xml – month: 04
  year: 2025
  text: 2025-April-26
  day: 26
PublicationDecade 2020
PublicationTitle Proceedings / International Conference on Software Engineering
PublicationTitleAbbrev ICSE
PublicationYear 2025
Publisher IEEE
Publisher_xml – name: IEEE
SSID ssj0006499
Score 2.3099954
Snippet Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation...
SourceID ieee
SourceType Publisher
StartPage 2587
SubjectTerms Code Generation
Codes
Complexity theory
Correlation
Empirical Study
Encoding
Large language models
Semantics
Software engineering
Software reliability
Syntactics
Taxonomy
Title Towards Understanding the Characteristics of Code Generation Errors Made by Large Language Models
URI https://ieeexplore.ieee.org/document/11029951
WOSCitedRecordID wos001538318100202&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV07a8MwEBZt6NApfbj0jYauaiy_PZuEFtoQaALZgiydoFDiYieB_vveyU6aDh26CKPFoEP30n3fx9hDHhoZWN8KUqQUxD8issgGgkTeMdwaDKnaiU2k43E2n-eTDqzusDAA4IbP4JE-3Vu-qfSaWmUDDFXoPQkwfZimSQvW2rndBHP3Dhsn_XzwXLwN4ziMUqwBA-qbSGJ-3FNQcQFk1P_nr0-Y9wPF45NdkDllB7A8Y_2tFgPvruY5U1M3_9rw2T5ahWN6x4vfnMy8sryoDPCWcZoMw4d1XdUNf1W4XX7xFxoPx7VtZXLSS_toPDYbDafFk-jkE4TCImUllMzKREqliEIsi20EGYQaS1LfKGksEBWNBguY8xhinbcqKnNlcxP7OvBVGF6w3rJawiXjWoLNQ60w2zKR0VCquMxKo9AZYMWj0yvm0ZEtPluGjMX2tK7_2L9hx2QVepUJklvWW9VruGNHerN6b-p7Z9dvDICkqg
linkProvider IEEE
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV3Na8IwFA_DDbaT-3Dseznsmtm0qbbnoihTEabgTdLkBQbDjlYH--_3XludO-ywSyi5FPLI-8r7_X6MPcWBlb7znCBFSkH8IyJSzhck8o7h1mJINaXYRHcyiRaLeFqD1UssDACUw2fwTJ_lW77NzIZaZW0MVeg9CTB9GCrlexVca-d4O5i91-g46cXtYfLaC8NAdbEK9KlzIon7cU9DpQwh_eY_f37KWj9gPD7dhZkzdgCrc9bcqjHw-nJeMD0rJ2ALPt_Hq3BM8Hjym5WZZ44nmQVecU6TaXgvz7O84GON2-kXH9GAOK5VM5OTYtp70WLzfm-WDEQtoCA0lilroWWUdqTUmkjEotApiCAwWJR6VkvrgMhoDDjArMcS77zTKo21i23oGd_TQXDJGqtsBVeMGwkuDozGfMsqayDVYRqlVqM7wJrHdK9Zi45s-VFxZCy3p3Xzx_4jOx7MxqPlaDh5uWUnZCF6o_E7d6yxzjdwz47M5_qtyB9KG38DyGSn8Q
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Abook&rft.genre=proceeding&rft.title=Proceedings+%2F+International+Conference+on+Software+Engineering&rft.atitle=Towards+Understanding+the+Characteristics+of+Code+Generation+Errors+Made+by+Large+Language+Models&rft.au=Wang%2C+Zhijie&rft.au=Zhou%2C+Zijie&rft.au=Song%2C+Da&rft.au=Huang%2C+Yuheng&rft.date=2025-04-26&rft.pub=IEEE&rft.eissn=1558-1225&rft.spage=2587&rft.epage=2599&rft_id=info:doi/10.1109%2FICSE55347.2025.00180&rft.externalDocID=11029951