paper_type.json (257B)
1 { 2 "paper_type": "empirical", 3 "reason": "Primary contribution is experimental findings about types and causes of LLM code generation mistakes through systematic analysis on existing benchmarks, with quantitative results (F1 scores, error percentages)." 4 }