tomhu/RL4TG-Qwen2.5-3B-OPD-7B-Teacher

TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 16, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The tomhu/RL4TG-Qwen2.5-3B-OPD-7B-Teacher model is a 3.1 billion parameter language model based on Qwen2.5-3B-Instruct, specifically fine-tuned for Java unit-test generation. It utilizes Online Policy Distillation (OPD) with Qwen2.5-Coder-7B-Instruct as the teacher model. This model is optimized for research in generating unit tests for Java projects, trained on the Defects4J project-disjoint dataset. It supports a 32768 token context length and uses the Qwen2.5 chat template.

Loading preview...

Model Overview

This model, tomhu/RL4TG-Qwen2.5-3B-OPD-7B-Teacher, is a specialized 3.1 billion parameter language model built upon the Qwen/Qwen2.5-3B-Instruct base. Its primary distinction lies in its fine-tuning for Java unit-test generation using an advanced technique called Online Policy Distillation (OPD). The distillation process leverages Qwen/Qwen2.5-Coder-7B-Instruct as a powerful teacher model, guiding the smaller 3B model to excel in this specific task.

Key Capabilities

  • Java Unit-Test Generation: Specifically trained and optimized to generate unit tests for Java codebases.
  • Online Policy Distillation (OPD): Employs a sophisticated training method to transfer knowledge from a larger, more capable teacher model.
  • Qwen2.5 Architecture: Benefits from the robust architecture of the Qwen2.5 series, supporting a 32768 token context length.
  • Research-Oriented: Intended for academic and research purposes in the domain of automated software testing and code generation.

Good For

  • Automated Test Generation Research: Ideal for researchers exploring methods to automatically create unit tests for Java applications.
  • Code Quality Improvement: Can be integrated into research pipelines aimed at enhancing software reliability through comprehensive testing.
  • Defects4J Dataset Applications: Trained on the Defects4J project-disjoint dataset, making it particularly relevant for studies involving this benchmark.

Users should note that generated tests require compilation and execution within the target project's build environment to validate their correctness and efficacy.

Finetunes
1 models