FRPO/qwen3-1.7b-a2_noclip-k1-cNone-clipInf-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-a2_noclip-k1-cNone-clipInf-mb4-eta100-bs256x5-n2 is a 2 billion parameter language model based on the Qwen3-1.7B architecture. This checkpoint is a result of RL fine-tuning experiments using the KL-in-LLM-RL / FRPO method, trained with the verl framework. It provides a specific iteration of reinforcement learning applied to the Qwen3-1.7B base model. This model is primarily suited for researchers and developers exploring the effects of RL fine-tuning on language models.

Loading preview...

Model Overview

This repository hosts qwen3-1.7b-a2_noclip-k1-cNone-clipInf-mb4-eta100-bs256x5-n2, a 2 billion parameter language model derived from the Qwen/Qwen3-1.7B base model. It represents a specific checkpoint from the KL-in-LLM-RL / FRPO experimental series, which focuses on reinforcement learning (RL) fine-tuning techniques. The training was conducted using the verl framework.

Key Characteristics

  • Base Model: Qwen3-1.7B, a compact yet capable foundation model.
  • Fine-tuning Method: Utilizes the KL-in-LLM-RL / FRPO approach, indicating a focus on specific reinforcement learning strategies.
  • Training Framework: Implemented with verl, a reinforcement learning library.
  • Weights: Provided in fp32 safetensors format, preserving the exact state as saved by the trainer without any post-processing.
  • Configuration: The specific run configuration details are encoded directly within the repository name, offering transparency into its experimental setup.

Intended Use Cases

This model is particularly relevant for:

  • RL Research: Researchers investigating the impact and effectiveness of KL-in-LLM-RL / FRPO fine-tuning methods on language models.
  • Experimental Development: Developers looking to build upon or analyze models that have undergone specific reinforcement learning optimizations.
  • Comparative Studies: As a benchmark or comparison point for other RL-tuned models or different fine-tuning strategies applied to the Qwen3-1.7B base.

Finetunes
1 models