MLC demonstrates that line-level bug localization at full-file context can match expensive agentic pipelines in accuracy while cutting inference to a single generated token per file, directly addressing the cost and latency barriers the paper identifies as blocking practical verification of LLM-generated code.