Keepace.ai post-trains Qwen3.5-27B for exercise guidance
Original title: Keepace.ai: A Technical Report on a Large Language Model for Sports and Health
Internal expert rubrics report gains in safety, boundary compliance, output completeness, and scientific accuracy.
Keepace.ai is a domain LLM built by post-training Qwen3.5-27B for Chinese-language exercise guidance. It targets workout plan generation, sports-and-health knowledge QA, and personal sports data analysis. The report describes the design, data pipeline, and a safety-first evaluation framework. In internal rubric evaluations with domain experts, the authors say it beats the base model and larger general-purpose baselines, especially on exercise safety and scientific accuracy.
Unconfirmed: Outperformance is from the authors' internal evaluations, not an independent benchmark.
- The base model is Qwen3.5-27B.
- Tasks are plans, knowledge QA, and personal sports data analysis.
- The authors say the domain still lacks curated corpora and rigorous benchmarks.
- Gains are largest on safety, boundaries, structured output, and accuracy.
From the original article
1 min read at arxiv.orgLarge language models hold promise for delivering safe, scientifically grounded, and personalized exercise guidance at scale, yet the sports-and-health domain still lacks curated training corpora and rigorous evaluation benchmarks. Keepace. ai is a domain-specialized LLM built on Qwen3.
By Keepace. ai Team · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.