cjiao/goldengoose-gumbel_combined_random_seed1-25grp

TEXT GENERATIONConcurrency Cost:1Model Size:1.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:May 31, 2026Architecture:Transformer Cold

The cjiao/goldengoose-gumbel_combined_random_seed1-25grp is a 1.5 billion parameter instruction-tuned causal language model, fine-tuned from Qwen/Qwen2.5-1.5B-Instruct. It was trained using the GRPO method, as introduced in the DeepSeekMath paper, which suggests an optimization for mathematical reasoning. This model is primarily designed for general text generation tasks, leveraging its instruction-tuned base and specialized training approach.

Loading preview...

Model Overview

The cjiao/goldengoose-gumbel_combined_random_seed1-25grp is a 1.5 billion parameter language model, fine-tuned from the Qwen/Qwen2.5-1.5B-Instruct base model. It leverages the Transformer Reinforcement Learning (TRL) framework for its training process.

Key Differentiator: GRPO Training

A significant aspect of this model's development is its training with GRPO (Gumbel-softmax Reinforcement Learning with Policy Optimization). This method, detailed in the paper "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models" (arXiv:2402.03300), suggests an approach to enhance reasoning capabilities, particularly in mathematical contexts. While the base model is instruction-tuned for general tasks, the application of GRPO implies a potential focus on improving logical and reasoning-based responses.

Capabilities

  • Instruction Following: Inherits instruction-following capabilities from its Qwen2.5-1.5B-Instruct base.
  • Text Generation: Capable of generating coherent and contextually relevant text based on prompts.
  • Potential for Enhanced Reasoning: The GRPO training method suggests an optimization for tasks requiring more structured or mathematical reasoning, though specific benchmarks are not provided.

Use Cases

This model is suitable for various text generation applications where a compact yet capable instruction-tuned model is desired. Its GRPO training might make it particularly interesting for tasks that benefit from improved logical consistency or problem-solving, such as:

  • General conversational AI
  • Content creation
  • Question answering
  • Tasks requiring structured output or reasoning, where the GRPO method's benefits could be observed.

Finetunes
1 models