Using Probabilistic Programs to Train Inductive Reasoning in Large Language Models
Quick Answer
This paper shows that The Program-based Posterior Training (PPT) method enhances inductive reasoning in Large Language Models (LLMs) by generating 10,000 diverse scenarios for fine-tuning, leading to improved accuracy and alignment with human judgments.
Quick Take
This approach demonstrates significant gains in estimation and calibration over traditional methods, indicating a deeper understanding of uncertainty in .
Key Points
- PPT fine-tunes LLMs using probabilistic programs for inductive reasoning.
- 10,000 scenarios generated to improve model performance on held-out tasks.
- Significant accuracy gains observed in estimation and human alignment.
- Raw calibration improvements exceed those from post-hoc temperature scaling.
- PPT shows promise for reliable approximate inductive inference in LLMs.
Paper Resources
Source Excerpt
arXiv:2606. 09856v1 Announce Type: new Abstract: Post-training (LLMs) for reasoning typically focuses on deductive tasks such as mathematics and coding where correctness is verifiable. Yet, many real-world reasoning problems are inductive: agents must infer uncertain beliefs from sparse, ambiguous observations.
There are challenges to using standard fine-tuning methods for inductive reasoning, including difficulties in curating large-scale, high-quality labeled datasets and in handling targets that are inherently distributional. …
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.