Dataset for Master's thesis: Agent-based Test Generation with Large Language Models and Stryker
收藏资源简介:
Run data for a master's thesis evaluating LLM-based JavaScript unit-test generation with mutation analysis. Contains: (1) TestPilot runs for Codex cushman, GPT-3.5-turbo, and StarCoder (10 runs each, derived from the TestPilot artifact, doi:10.6084/m9.figshare.23653371, CC BY 4.0) augmented with StrykerJS mutation reports (2) two fresh TestPilot runs using GPT-5.4 (500/1000 output-token budgets) including per-test coverage data(3) full run trees of a custom mutation-guided agentic test generator (agentic-stryker) at commit 63be77f, including LLM transcripts and per-run provenance manifests: two temperature-1 full+residual replicates, one temperature-0 run, two hidden-feedback ablation runs, and a March-2026 pilot. All mutation reports are the post-idempotency-correction state. See ARCHIVE_README.md for the tarball-to-layout map and the modification statemen. The analysis pipeline that turns this data into the thesis tables and figures is at https://github.com/eefscheef/agentic-stryker-tables.



