arXiv:2606.09856v1 Announce Type: cross 
Abstract: Post-training Large Language Models (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. In this work, we introduce a novel approach, called Program-based Posterior Training (PPT), to address these limitations: we use an LLM to generate diverse open-world scenarios as probabilistic programs, run probabilistic inference to produce distributional target responses to queries, and then fine-tune on these probabilistic soft labels. Using this approach, we fine-tune LLMs on 10,000 programmatically generated scenarios and evaluate on held-out motifs, human-labeled judgments, and external benchmarks. Overall, PPT substantially improves estimation accuracy on held-out inductive tasks, increases alignment with human judgments, and transfers to external benchmarks for estimation and calibration. Additionally, the gains in raw calibration are not subsumed by post-hoc temperature scaling, showing that the models have more deeply internalized uncertainty compared to output rescaling. Together, these results suggest that probabilistic-program-mediated fine-tuning is a promising approach for post-training LLMs to reliably perform approximate inductive inference.

تم تقديم نهج جديد يسمى التدريب البعدي القائم على البرامج (PPT) لتعزيز التفكير الاستقرائي في نماذج اللغة الكبيرة (LLMs). يتناول هذا الأسلوب التحديات المتعلقة بضبط نماذج LLM من خلال إنشاء سيناريوهات متنوعة كبرامج احتمالية وضبط الاستجابات المستهدفة الناتجة.

Se ha introducido un nuevo enfoque llamado Entrenamiento Posterior Basado en Programas (PPT) para mejorar el razonamiento inductivo en Modelos de Lenguaje de Gran Escala (LLMs). Este método aborda los desafíos del ajuste fino de los LLMs generando escenarios diversos como programas probabilísticos y ajustando las respuestas objetivo resultantes.

Une nouvelle approche appelée Formation Postérieure Basée sur des Programmes (PPT) a été introduite pour améliorer le raisonnement inductif dans les Modèles de Langage de Grande Taille (LLMs). Cette méthode répond aux défis de l'affinage des LLMs en générant des scénarios divers en tant que programmes probabilistes et en affinant les réponses cibles distributionnelles qui en résultent.

A novel approach called Program-based Posterior Training (PPT) has been introduced to enhance inductive reasoning in Large Language Models (LLMs). This method addresses challenges in fine-tuning LLMs by generating diverse scenarios as probabilistic programs and fine-tuning on the resulting distributional target responses.