HYU-NLP-EVAL/20260914_rar-medicine_static_qwen3-4b_matched_impl-verl_s11_step-006

TEXT GENERATIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 30, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

HYU-NLP-EVAL/20260914_rar-medicine_static_qwen3-4b_matched_impl-verl_s11_step-006 is a 4-billion parameter causal language model based on Qwen3-4B-Instruct, developed by HYU-NLP-EVAL. This model is an intermediate research checkpoint from a matched static-rubric GRPO experiment, specifically fine-tuned on the RaR-Medicine dataset for medical domain applications. It is optimized for tasks related to medicine, leveraging a static reward source and GRPO for policy updates, making it suitable for research in medical NLP.

Loading preview...

Overview

This model, HYU-NLP-EVAL/20260914_rar-medicine_static_qwen3-4b_matched_impl-verl_s11_step-006, is a 4-billion parameter policy derived from Qwen/Qwen3-4B-Instruct-2507. It represents an intermediate checkpoint from a research experiment focusing on static-rubric GRPO (Global Reward Policy Optimization) within the medical domain. The model has undergone 6 global optimizer updates as part of a larger research initiative.

Key Capabilities

  • Medical Domain Specialization: Fine-tuned using the RaR-Medicine dataset (1,500 prompts), making it specialized for medical text understanding and generation tasks.
  • GRPO Optimization: Utilizes a static reward source (rar_static_r0_only) and the GRPO method for policy refinement, indicating a focus on optimizing responses based on predefined rubrics.
  • Research Checkpoint: This is an intermediate research checkpoint, providing insights into the progression of the GRPO training process for medical NLP.

Good for

  • Medical NLP Research: Ideal for researchers exploring policy optimization techniques like GRPO in the medical domain.
  • Experimentation with Static Rubrics: Useful for studying the effects of static reward signals on language model performance in specialized fields.
  • Understanding GRPO Progression: Provides a snapshot of the model's capabilities after a specific number of GRPO updates, allowing for analysis of training dynamics.

Note: This is a research checkpoint and is not intended for clinical use. No medical capability or safety claims are made.

Finetunes
1 models