{"id":{"repo_id":"vilnius","oai_identifier":"oai:vu.lt:elaba:308337250"},"canonical_url":"https://search.ndltd.org/etd/vilnius/oai:vu.lt:elaba:308337250","repository":{"repo_id":"vilnius","name":"Vilnius University","base_url":"https://epublications.vu.lt/oai"},"display":{"title":"Agentų didžiųjų kalbos modelių taikymas integraciniam testavimui /","abstract":"This thesis examines the application of generative artificial intelligence based on large language models for creating integration tests. The properties of high&#8208;quality integration tests and test evaluation metrics are analysed. An overview of popular generative artificial intelligence tools used for integration test generation is provided. An experimental study is carried out, evaluating the quality of generative artificial intelligence tools with different prompts. GenAI tools are analysed. ”GitHub Copilot” and ”Claude Code” are selected for the practical experiment. ”Pro” plan access is used for the GenAI tools. During the experiment, integration tests covering 80 to 90% of the API layer are generated for the ”PSK&#8208;Project” and ”POS&#8208;System” projects. The practical part is divided into separate experiments according to the selected parameters: three different prompts and two generative artificial intelligence tools. The GenAI tools are given three prompts: a simple prompt, a prompt with requirements specifying what the test should look like, and a prompt with requirements and an example. Each experiment is repeated three times. The quality of the GenAI tools is evaluated based on their generated results through quantitative and qualitative analyses. In the quantitative analysis, the integration tests are evaluated according to the selected metrics: line coverage, branch coverage, average test execution time and mutation testing score. During the qualitative analysis, the compliance of the generated integration tests with the established requirements is evaluated. The experiment found that additional contextual information in the prompt significantly improves the quality of the generated integration tests. Adding structural requirements leads to the appropriate test type, which in this case is integration test. Providing an example results in a more realistic testing infrastructure and more thorough verification of business logic. The investigated GenAI tools differ not in the maximum achievable quality of integration tests, but in the stability and reproducibility of the results. ”Claude Code” consistently generates high&#8208;quality tests, whereas ”GitHub Copilot” exhibits greater variability in results and a stronger dependence on the project under test. Quantitative indicators alone do not reflect test quality, and therefore qualitative analysis based on requirements is essential for reliable evaluation.","abstract_html":"This thesis examines the application of generative artificial intelligence based on large language models for creating integration tests. The properties of high&amp;#8208;quality integration tests and test evaluation metrics are analysed. An overview of popular generative artificial intelligence tools used for integration test generation is provided. An experimental study is carried out, evaluating the quality of generative artificial intelligence tools with different prompts. GenAI tools are analysed. ”GitHub Copilot” and ”Claude Code” are selected for the practical experiment. ”Pro” plan access is used for the GenAI tools. During the experiment, integration tests covering 80 to 90% of the API layer are generated for the ”PSK&amp;#8208;Project” and ”POS&amp;#8208;System” projects. The practical part is divided into separate experiments according to the selected parameters: three different prompts and two generative artificial intelligence tools. The GenAI tools are given three prompts: a simple prompt, a prompt with requirements specifying what the test should look like, and a prompt with requirements and an example. Each experiment is repeated three times. The quality of the GenAI tools is evaluated based on their generated results through quantitative and qualitative analyses. In the quantitative analysis, the integration tests are evaluated according to the selected metrics: line coverage, branch coverage, average test execution time and mutation testing score. During the qualitative analysis, the compliance of the generated integration tests with the established requirements is evaluated. The experiment found that additional contextual information in the prompt significantly improves the quality of the generated integration tests. Adding structural requirements leads to the appropriate test type, which in this case is integration test. Providing an example results in a more realistic testing infrastructure and more thorough verification of business logic. The investigated GenAI tools differ not in the maximum achievable quality of integration tests, but in the stability and reproducibility of the results. ”Claude Code” consistently generates high&amp;#8208;quality tests, whereas ”GitHub Copilot” exhibits greater variability in results and a stronger dependence on the project under test. Quantitative indicators alone do not reflect test quality, and therefore qualitative analysis based on requirements is essential for reliable evaluation.","abstract_has_math":false,"creators":["Kasperavičiūtė, Brigita,"],"institution":"Institutional Repository of Vilnius University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-19T10:12:59Z","subjects":[],"languages":["lit"],"rights":["info:eu-repo/semantics/openAccess"],"rights_urls":[],"identifier_entries":[],"person_identifier_entries":[]},"links":{"outbound_url":"https://repository.vu.lt/VU:ELABAETD308337250&prefLang=en_US","outbound_label":"Repository record","outbound_source":"dc:identifier","record_url":"https://repository.vu.lt/VU:ELABAETD308337250&prefLang=en_US","record_label":"Repository record","record_kind":"record","record_source":"dc:identifier","download_url":null,"download_label":null,"download_source":null,"access_status":"open","embargo_until":null,"access_note":null},"source_record":{"url":"https://epublications.vu.lt/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Avu.lt%3Aelaba%3A308337250","prefix":"oai_dc"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Kasperavičiūtė, Brigita,"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026"]},{"key":"dc:publisher","label":"Institution","values":["Institutional Repository of Vilnius University"]},{"key":"dc:relation","label":"Dc Relation","values":["https://epublications.vu.lt/object/elaba:308337250/308337250.pdf"]},{"key":"dc:type","label":"Dc Type","values":["info:eu-repo/semantics/bachelorThesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language","label":"Dc Language","values":["lit"]},{"key":"dc:rights","label":"Dc Rights","values":["info:eu-repo/semantics/openAccess"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://repository.vu.lt/VU:ELABAETD308337250&prefLang=en_US"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["This thesis examines the application of generative artificial intelligence based on large language models for creating integration tests. The properties of high&#8208;quality integration tests and test evaluation metrics are analysed. An overview of popular generative artificial intelligence tools used for integration test generation is provided. An experimental study is carried out, evaluating the quality of generative artificial intelligence tools with different prompts. GenAI tools are analysed. ”GitHub Copilot” and ”Claude Code” are selected for the practical experiment. ”Pro” plan access is used for the GenAI tools. During the experiment, integration tests covering 80 to 90% of the API layer are generated for the ”PSK&#8208;Project” and ”POS&#8208;System” projects. The practical part is divided into separate experiments according to the selected parameters: three different prompts and two generative artificial intelligence tools. The GenAI tools are given three prompts: a simple prompt, a prompt with requirements specifying what the test should look like, and a prompt with requirements and an example. Each experiment is repeated three times. The quality of the GenAI tools is evaluated based on their generated results through quantitative and qualitative analyses. In the quantitative analysis, the integration tests are evaluated according to the selected metrics: line coverage, branch coverage, average test execution time and mutation testing score. During the qualitative analysis, the compliance of the generated integration tests with the established requirements is evaluated. The experiment found that additional contextual information in the prompt significantly improves the quality of the generated integration tests. Adding structural requirements leads to the appropriate test type, which in this case is integration test. Providing an example results in a more realistic testing infrastructure and more thorough verification of business logic. The investigated GenAI tools differ not in the maximum achievable quality of integration tests, but in the stability and reproducibility of the results. ”Claude Code” consistently generates high&#8208;quality tests, whereas ”GitHub Copilot” exhibits greater variability in results and a stronger dependence on the project under test. Quantitative indicators alone do not reflect test quality, and therefore qualitative analysis based on requirements is essential for reliable evaluation."]},{"key":"dc:format","label":"Dc Format","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Agentų didžiųjų kalbos modelių taikymas integraciniam testavimui /","Application of agent large language models for integration testing."]}]}],"canonical_facts":{"dc:creator":["Kasperavičiūtė, Brigita,"],"dc:date":["2026"],"dc:description":["This thesis examines the application of generative artificial intelligence based on large language models for creating integration tests. The properties of high&#8208;quality integration tests and test evaluation metrics are analysed. An overview of popular generative artificial intelligence tools used for integration test generation is provided. An experimental study is carried out, evaluating the quality of generative artificial intelligence tools with different prompts. GenAI tools are analysed. ”GitHub Copilot” and ”Claude Code” are selected for the practical experiment. ”Pro” plan access is used for the GenAI tools. During the experiment, integration tests covering 80 to 90% of the API layer are generated for the ”PSK&#8208;Project” and ”POS&#8208;System” projects. The practical part is divided into separate experiments according to the selected parameters: three different prompts and two generative artificial intelligence tools. The GenAI tools are given three prompts: a simple prompt, a prompt with requirements specifying what the test should look like, and a prompt with requirements and an example. Each experiment is repeated three times. The quality of the GenAI tools is evaluated based on their generated results through quantitative and qualitative analyses. In the quantitative analysis, the integration tests are evaluated according to the selected metrics: line coverage, branch coverage, average test execution time and mutation testing score. During the qualitative analysis, the compliance of the generated integration tests with the established requirements is evaluated. The experiment found that additional contextual information in the prompt significantly improves the quality of the generated integration tests. Adding structural requirements leads to the appropriate test type, which in this case is integration test. Providing an example results in a more realistic testing infrastructure and more thorough verification of business logic. The investigated GenAI tools differ not in the maximum achievable quality of integration tests, but in the stability and reproducibility of the results. ”Claude Code” consistently generates high&#8208;quality tests, whereas ”GitHub Copilot” exhibits greater variability in results and a stronger dependence on the project under test. Quantitative indicators alone do not reflect test quality, and therefore qualitative analysis based on requirements is essential for reliable evaluation."],"dc:format":["application/pdf"],"dc:identifier":["https://repository.vu.lt/VU:ELABAETD308337250&prefLang=en_US"],"dc:language":["lit"],"dc:publisher":["Institutional Repository of Vilnius University"],"dc:relation":["https://epublications.vu.lt/object/elaba:308337250/308337250.pdf"],"dc:rights":["info:eu-repo/semantics/openAccess"],"dc:title":["Agentų didžiųjų kalbos modelių taikymas integraciniam testavimui /","Application of agent large language models for integration testing."],"dc:type":["info:eu-repo/semantics/bachelorThesis"]},"updated_at":"2026-11-19T10:12:59Z"}