This repository contains the reproducibility artifact for:
A Survey of LLM-based Automated Program Repair: Taxonomies, Design Paradigms, and Applications
Start from artifact/README.md.
The main auditable files for the current 66-system corpus are:
artifact/search_keywords_and_filters.mdartifact/screening_transparency.mdartifact/selection_reference_474_final_adjudicated.csvartifact/selection_reference_474_final_adjudicated_summary.jsonartifact/screening_agreement_labels_474.csvartifact/compute_screening_agreement.pyartifact/taxonomy_assignment_audit.csvartifact/taxonomy_independent_pair_66_audit.csvartifact/taxonomy_independent_pair_66_agreement_summary.mdartifact/scenario_assignment_audit.csvartifact/benchmark_protocol_summary.mdartifact/version_status_audit.mdartifact/reproduce_screening_counts.pyartifact/screening_count_reproduction_2026-05-01.jsonremote_results/pipeline.pyremote_results/stage4.jsonlremote_results/stage5.jsonlremote_results/stage6.jsonl
- The public artifact focuses on auditable metadata, screening decisions, taxonomy assignments, benchmark summaries, and the released retrieval/filtering stages.
- Local submission-package folders, draft response files, temporary annotation handoff files, and generated text caches are excluded from Git tracking.
- The retained-paper full-text PDFs are included as auditable source material. The structured audit data, reproduction scripts, and table-to-file mappings under
artifact/remain the authoritative interface for reproducing manuscript counts and labels.