1

Show HN: Training-free improvement of Pareto frontiers for reasoning LLMs