HYU-NLP-EVAL/qwen3-4b-rar-medicine-onlinerubrics-seed11-step-040
HYU-NLP-EVAL/qwen3-4b-rar-medicine-onlinerubrics-seed11-step-040 is an intermediate policy checkpoint from dynamic OnlineRubrics-Every GRPO training, distinct from static-rubric GRPO. Based on the Qwen3-4B-Instruct-2507 model, this 4 billion parameter model with a 32768 token context length is specifically a historical policy state used by the Phase-1 audit. It is intended for research use only, with no downstream medical capability or safety claims, and is not validated for clinical decision-making.
Loading preview...
Model Overview
This model, HYU-NLP-EVAL/qwen3-4b-rar-medicine-onlinerubrics-seed11-step-040, is an intermediate policy checkpoint derived from dynamic OnlineRubrics-Every GRPO training. It is distinct from static-rubric GRPO and is based on the Qwen/Qwen3-4B-Instruct-2507 model, with thinking capabilities disabled.
Key Characteristics
- Training Origin: Represents a specific historical policy state from a Phase-1 audit of dynamic OnlineRubrics-Every GRPO training.
- Base Model: Built upon the Qwen3-4B-Instruct-2507 architecture.
- Parameter Count: 4 billion parameters.
- Context Length: Supports a context length of 32768 tokens.
- Research Focus: Explicitly designated for research use only.
Important Considerations
- No Medical Claims: The developers explicitly state that no downstream medical capability or safety claim is made.
- Not for Clinical Use: This model is not validated for clinical decision-making.
- Checkpoint Type: The provided files are veRL-exported Hugging Face inference models (BF16), with the original FSDP parameter checkpoint preserved separately for precision reasons. Optimizer state, training data, responses, rubrics, infrastructure configuration, and credentials are not included.