![[2602.15449] TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15449] TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models
Summary
The TAROT framework enhances code generation in large language models by implementing a test-driven, capability-adaptive reinforcement fine-tuning approach, optimizing curriculum design for improved code quality.
Why It Matters
As large language models revolutionize coding, ensuring robust and efficient code generation becomes crucial. TAROT addresses the challenges of imbalanced reward signals in training, proposing a structured curriculum that adapts to model capabilities, fostering better performance and reliability in code generation tasks.
Key Takeaways
- TAROT introduces a four-tier test suite for structured curriculum design.
- The framework adapts training based on the model's capability, optimizing learning outcomes.
- Less capable models benefit from an easy-to-hard progression, while more competent models perform better with a hard-first approach.
- TAROT enhances the functional correctness and robustness of generated code.
- The method is reproducible, promoting further research and collaboration in the AI community.
Computer Science > Computation and Language arXiv:2602.15449 (cs) [Submitted on 17 Feb 2026] Title:TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models Authors:Chansung Park, Juyong Jiang, Fan Wang, Sayak Paul, Jiasi Shen, Jing Tang, Jianguo Li View a PDF of the paper titled TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models, by Chansung Park and 6 other authors View PDF HTML (experimental) Abstract:Large Language Models (LLMs) are changing the coding paradigm, known as vibe coding, yet synthesizing algorithmically sophisticated and robust code still remains a critical challenge. Incentivizing the deep reasoning capabilities of LLMs is essential to overcoming this hurdle. Reinforcement Fine-Tuning (RFT) has emerged as a promising strategy to address this need. However, most existing approaches overlook the heterogeneous difficulty and granularity inherent in test cases, leading to an imbalanced distribution of reward signals and consequently biased gradient updates during training. To address this, we propose Test-driven and cApability-adaptive cuRriculum reinfOrcement fine-Tuning (TAROT). TAROT systematically constructs, for each problem, a four-tier test suite (basic, intermediate, complex, edge), providing a controlled difficulty landscape for curriculum design and evaluation. Crucially, TAROT decouples curriculum progress...