Enhancing Fitness Intelligence through Domain-Specific LLM Post-Training

2026-07-02

AI Summary

This work addresses the limited reliability of general-purpose large language models in scientific fitness guidance due to insufficient domain-specific knowledge. To overcome this, we propose the FitOne series of models (8B/32B), built upon Qwen3 and trained via a three-stage post-training paradigm—comprising continued pretraining, supervised fine-tuning, and reinforcement learning—augmented with high-quality, expert-curated fitness knowledge engineering data. This approach substantially enhances domain expertise while preserving strong general capabilities. Experimental results demonstrate that FitOne-8B and FitOne-32B achieve average performance gains of 10.09%/9.29% on the ACSM-EP exam and 12.73%/7.01% on the NSCA-CSCS certification test, respectively, thereby validating the effectiveness and generalization ability of the proposed methodology.

Abstract

Scientific Fitness Coaching (SFC) is typically delivered by human professionals, making it costly and inaccessible to many. While recent advances in Large Language Models (LLMs) show considerable promise for more inclusive fitness coaching, directly deploying prevailing general-purpose LLMs in SFC reveals critical limitations. These models often lack sufficient domain-specific knowledge integration, leading to weak performance on complex SFC scenarios. In this paper, we introduce FitOne, a series of fitness LLMs (with 8B and 32B parameters) designed to improve reliability and domain specialization for SFC applications. Built upon the Qwen3 foundation models, FitOne is developed through a three-stage post-training pipeline consisting of continual pre-training, supervised fine-tuning, and reinforcement learning, using large-scale, high-quality datasets derived from rigorous knowledge engineering. We conduct comprehensive evaluations of FitOne on professional fitness certification exams, including ACSM-EP and NSCA-CSCS, as well as general capabilities such as knowledge reasoning and instruction following. Experimental results show that, while retaining strong general capabilities, FitOne-8B/32B achieves average improvements of up to 10.09%/9.29% and 12.73%/7.01% on the ACSM-EP and NSCA-CSCS exams, respectively, compared with the Qwen3 base models. Furthermore, in-depth ablation studies confirm the necessity of each training stage, highlighting the pipeline's effectiveness in balancing domain expertise enhancement with general ability retention. We believe this research advances LLM systems toward more reliable fitness intelligence and will inspire future research on developing domain-specific LLMs.

Problem

• Scientific Fitness Coaching
• Large Language Models
• Domain-Specific Knowledge
• Fitness Intelligence
• LLM Limitations

Innovation

• domain-specific LLM
• post-training pipeline
• fitness intelligence
• knowledge engineering
• reinforcement learning