nilgeoutim/RLVR-hotpot

TEXT GENERATIONConcurrent Unit Cost:1Model Size:3.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 3, 2026Architecture:Transformer Featherless Exclusive Cold

nilgeoutim/RLVR-hotpot is a 3.1 billion parameter causal language model fine-tuned from Qwen/Qwen2.5-3B. Developed by nilgeoutim, this model utilizes the GRPO training method, as introduced in the DeepSeekMath paper, to enhance mathematical reasoning capabilities. With a context length of 32768 tokens, it is optimized for tasks requiring robust reasoning and problem-solving, particularly in areas benefiting from advanced mathematical understanding.

Loading preview...

Model Overview

nilgeoutim/RLVR-hotpot is a 3.1 billion parameter language model, fine-tuned from the Qwen/Qwen2.5-3B base model. It leverages the GRPO (Gradient-based Reward Policy Optimization) training method, a technique highlighted in the DeepSeekMath paper, which focuses on pushing the limits of mathematical reasoning in open language models. This fine-tuning process was conducted using the TRL framework.

Key Capabilities

  • Enhanced Reasoning: Optimized for tasks requiring logical deduction and problem-solving, particularly in mathematical contexts, due to its GRPO-based training.
  • Large Context Window: Supports a context length of 32768 tokens, allowing for processing and generating longer sequences of text.
  • Instruction Following: Capable of generating responses based on user prompts, as demonstrated by its quick start example.

Good For

  • Mathematical Reasoning Tasks: Ideal for applications where strong mathematical understanding and problem-solving are crucial.
  • Complex Query Handling: Suitable for scenarios requiring the model to process and respond to intricate questions or instructions.
  • Research and Development: Provides a foundation for further experimentation with GRPO-trained models in reasoning-intensive domains.