tomhu/RL4TG-Qwen2.5-3B-GRPO-1-Epoch
tomhu/RL4TG-Qwen2.5-3B-GRPO-1-Epoch is a 3.1 billion parameter language model based on the Qwen2.5 architecture, fine-tuned using Group Relative Policy Optimization (GRPO). This model is specifically designed and optimized for the generation of Java unit tests. It is intended for research purposes in the domain of automated test generation, particularly for mutation-eligible Defects4J samples.
Loading preview...
Model Overview
tomhu/RL4TG-Qwen2.5-3B-GRPO-1-Epoch is a specialized 3.1 billion parameter language model derived from the Qwen/Qwen2.5-3B-Instruct base model. It has been fine-tuned using the Group Relative Policy Optimization (GRPO) method over a single epoch (50 steps) with a global training seed of 42.
Key Capabilities
- Java Unit Test Generation: The model's primary function is to generate Java unit tests, specifically targeting mutation-eligible samples from the Defects4J dataset.
- Research Focus: It is explicitly intended for research in automated test generation, providing a checkpoint for further study in this area.
- Qwen2.5 Architecture: Leverages the Qwen2.5 chat template for prompt formatting, ensuring compatibility with the base model's instruction-following capabilities.
Intended Use
This model is designed for researchers and developers working on Java unit test automation. It can be used to explore and evaluate the effectiveness of RL-based fine-tuning for code generation tasks. Users should note that generated tests require compilation and execution within the target project's build environment to verify their functionality and correctness.
Limitations
The model is a research checkpoint (1 epoch) and is highly specialized for Java unit test generation. Its performance on other general-purpose language tasks or different programming languages is not its intended use case.