Using Probabilistic Programs to Train Inductive Reasoning in Large Language Models

In three linesNew Program-based Posterior Training (PPT) method to train LLMs for inductive reasoning. Uses LLM-generated probabilistic programs to create 10,000 training scenarios with distributional labels. Significantly improves accuracy, human alignment, and calibration independent of temperature scaling.

Summary generated by Claude — human-verified