HYU-NLP-EVAL/qwen3-1.7b-rar-medicine-static-r0-step-040

TEXT GENERATIONPricing:Input $0.32 / Cached $0.064 / Output $1.6Concurrent Unit Cost:1Model Size:2BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 3, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The HYU-NLP-EVAL/qwen3-1.7b-rar-medicine-static-r0-step-040 is a 1.7 billion parameter Qwen3-based causal language model, developed by HYU-NLP-EVAL. This specific checkpoint is a research artifact from a static-rubric discriminability-horizon experiment, fine-tuned using the GRPO algorithm with a frozen prompt-specific initial rubric (R0) in the medical domain. It is intended for studying reward saturation and static-rubric staleness during policy optimization, rather than for direct application as a medical device.

Loading preview...

Model Overview

This repository hosts a specific policy checkpoint, qwen3-1.7b-rar-medicine-static-r0-step-040, derived from the Qwen/Qwen3-1.7B base model. It is a research artifact from the HYU-NLP-EVAL group's static-rubric discriminability-horizon experiment, focusing on the medical domain.

Key Characteristics

  • Base Model: Qwen3-1.7B, a 1.7 billion parameter causal language model.
  • Training Algorithm: Fine-tuned using the GRPO (Generalized Reinforcement Policy Optimization) algorithm.
  • Reward Mechanism: Utilizes a frozen prompt-specific initial rubric (R0) for training.
  • Domain Focus: Specifically trained within the RaR Medicine domain, with this checkpoint representing step 40 of the optimization process.
  • Export Format: Provided in Hugging Face Transformers format, using BF16 safetensors, including model weights, configuration, tokenizer, and chat template.

Intended Use

This model is primarily a research artifact designed for:

  • Studying reward saturation in policy optimization.
  • Investigating static-rubric staleness during policy optimization.

Important Note: Medicine checkpoints, including this one, are not medical devices and should not be used as a substitute for professional medical advice. They are for research purposes only to understand the dynamics of reinforcement learning in specific domains.

Finetunes
1 models