luckeciano/Qwen-2.5-7B-GRPO-Adam-FisherMaskToken-1e-4-HessianMaskToken-0.01-v2_9494

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Sep 14, 2025Architecture:Transformer Featherless Exclusive Cold

The luckeciano/Qwen-2.5-7B-GRPO-Adam-FisherMaskToken-1e-4-HessianMaskToken-0.01-v2_9494 is a 7.6 billion parameter language model, fine-tuned from Qwen/Qwen2.5-Math-7B. It was trained using the GRPO method on the MATH-lighteval dataset, specializing it in mathematical reasoning. This model is designed for advanced mathematical problem-solving and complex reasoning tasks, leveraging its specialized training for enhanced accuracy in numerical and logical operations.

Loading preview...

Overview

This model, luckeciano/Qwen-2.5-7B-GRPO-Adam-FisherMaskToken-1e-4-HessianMaskToken-0.01-v2_9494, is a 7.6 billion parameter language model derived from the Qwen/Qwen2.5-Math-7B base model. It has been specifically fine-tuned using the GRPO (Generalized Reinforcement Learning with Policy Optimization) method, as introduced in the DeepSeekMath paper, to enhance its mathematical reasoning capabilities. The training utilized the DigitalLearningGmbH/MATH-lighteval dataset, making it particularly adept at handling complex mathematical problems.

Key Capabilities

  • Advanced Mathematical Reasoning: Specialized training on the MATH-lighteval dataset using GRPO significantly improves its ability to solve intricate mathematical problems.
  • Fine-tuned Performance: Leverages the robust architecture of Qwen 2.5 with targeted optimization for numerical and logical tasks.
  • GRPO Method: Incorporates a sophisticated training approach for improved stability and performance in reinforcement learning from human feedback (RLHF) contexts, particularly beneficial for reasoning tasks.

Good For

  • Mathematical Problem Solving: Ideal for applications requiring high accuracy in arithmetic, algebra, geometry, and other mathematical domains.
  • Research and Development: Suitable for researchers exploring advanced fine-tuning techniques and their impact on reasoning tasks.
  • Educational Tools: Can be integrated into platforms that assist with complex math homework or provide detailed step-by-step solutions.

Finetunes
1 models