Runs HLE-oriented benchmark reward ingestion and curriculum generation for capability-evolver. Use when the user asks to optimize Humanity's Last Exam score, ingest question-level benchmark results, prioritize easy-first queues, or request an immediate benchmark progress result.