DeepMath-Omn-1.5B: Advanced Mathematical Reasoning
DeepMath-Omn-1.5B is a 1.5 billion parameter model from zwhe99, fine-tuned via reinforcement learning on the specialized DeepMath-103K dataset. This model is engineered to excel in complex mathematical reasoning, addressing problems primarily from difficulty levels 5-9, which significantly surpasses the complexity of many existing datasets.
Key Capabilities & Features
- Challenging Mathematical Problems: Focuses on difficult problems across diverse domains like Algebra, Calculus, Number Theory, Geometry, Probability, and Discrete Mathematics.
- Novel Data Diversity: Trained on the unique DeepMath-103K dataset, which features novel problems and broad topic coverage, minimizing overlap with existing datasets.
- Rigorous Decontamination: The DeepMath-103K dataset underwent meticulous decontamination against common benchmarks to ensure fair model evaluation and prevent test set leakage.
- Rich Data Format: Each training sample includes the question, a reliably verifiable final answer, a numerical difficulty score, hierarchical topic classification, and three distinct reasoning paths (R1 Solutions) for robust training.
- State-of-the-Art Performance: Achieves strong results on challenging math benchmarks, demonstrating advanced mathematical reasoning abilities.
Ideal Use Cases
- Mathematical Problem Solving: Excels at solving complex mathematical problems, particularly those requiring advanced reasoning.
- Research in Mathematical AI: Useful for researchers exploring the boundaries of AI in mathematics, leveraging its challenging dataset and verifiable answers.
- Educational Tools: Can be integrated into systems requiring high-accuracy mathematical solutions and reasoning path generation.