FRPO/qwen3-1.7b-a8_klreward-krew-k1-coef1.0-mb4-eta100-bs256x5-n2

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Aug 15, 2026Architecture:Transformer Featherless Exclusive Cold

FRPO/qwen3-1.7b-a8_klreward-krew-k1-coef1.0-mb4-eta100-bs256x5-n2 is a 1.7 billion parameter language model based on the Qwen3 architecture, specifically fine-tuned using Reinforcement Learning (RL) with the KL-in-LLM-RL / FRPO experimental framework. Developed by FRPO, this model leverages the verl training library. It is designed for tasks benefiting from RL-based optimization, offering a specialized approach to language generation and understanding.

Loading preview...

Overview

This model, FRPO/qwen3-1.7b-a8_klreward-krew-k1-coef1.0-mb4-eta100-bs256x5-n2, is a 1.7 billion parameter language model derived from the Qwen/Qwen3-1.7B base model. It has undergone specific fine-tuning using Reinforcement Learning (RL) within the KL-in-LLM-RL / FRPO experimental framework, utilizing the verl training library.

Key Characteristics

  • Base Model: Qwen3-1.7B, providing a robust foundation for language tasks.
  • Fine-tuning Method: Employs Reinforcement Learning (RL) with a focus on KL-in-LLM-RL / FRPO experiments.
  • Checkpoint: The repository contains the global_step_200 checkpoint.
  • Weights: Stored in fp32 safetensors format, directly as saved by the trainer without post-processing.
  • Configuration: The specific run configuration is encoded within the model's repository name, detailing parameters like a8_klreward-krew-k1-coef1.0-mb4-eta100-bs256x5-n2.

Potential Use Cases

This model is particularly suited for research and development in:

  • RL-based Language Generation: Exploring the effects of KL-in-LLM-RL and FRPO techniques on language model performance.
  • Experimental AI: Investigating novel fine-tuning strategies for large language models.
  • Specialized Language Tasks: Applications where RL fine-tuning can yield specific behavioral improvements over standard supervised fine-tuning.