Fundamental Reasoning Paradigms Induce Out-of-Domain Generalization in Language Models
Preprint
Abstract
Deduction, induction, and abduction are fundamental reasoning paradigms, core for human logical thinking. Although improving Large Language Model (LLM) reasoning has attracted significant research efforts, the extent to which the fundamental paradigms induce generalization has yet to be systematically explored. In this study, we shed light on how the interplay between these core paradigms influences LLMs' reasoning behavior. To this end, we first collect a new dataset of reasoning trajectories from symbolic tasks, each targeting one of the three fundamental paradigms, to abstract from concrete world knowledge. Then, we investigate effective ways for inducing these skills into LLMs. We experiment with a battery of methods including simple fine-tuning, and more complex approaches to increase model depth, or transform a dense model to a mixture-of-experts. We comprehensively evaluate induced models on realistic out-of-domain tasks, that are entirely formulated in natural language and contain real-world knowledge. Our results reveal that our approach yields strong generalizability with substantial performance gains (up to $14.60$) across realistic tasks.Keywords
All Related Versions
This publication has 18 references indexed in Scilit:
- Datasets for large language models: a comprehensive surveyArtificial Intelligence Review, 2025
- LogiDynamics: Unraveling the Dynamics of Inductive, Abductive and Deductive Logical Inferences in LLM Reasoning2025
- ENGinius: A Bilingual LLM Optimized for Plant Construction Engineering2025
- Less, but Better: Efficient Multilingual Expansion for LLMs via Layer-wise Mixture-of-Experts2025
- FOLIO: Natural Language Reasoning with First-Order Logic2024
- LLaMA Pro: Progressive LLaMA with Block Expansion2024
- True Detective: A Deep Abductive Reasoning Benchmark Undoable for GPT-3 and Challenging for GPT-42023
- Abductive Commonsense Reasoning Exploiting Mutually Exclusive Explanations2023
- Transformers: State-of-the-Art Natural Language Processing2020
- WinoWhy: A Deep Diagnosis of Essential Commonsense Knowledge for Answering Winograd Schema Challenge2020