Abdine
/

qwen3-4b-medrect-mixed-r2

Reinforcement Learning

text-generation

text-generation-inference

Model card Files Files and versions

Abdine/qwen3-4b-medrect-mixed-r2

Round-1 MedSeRL self-play actor exported from VERL training.

Base model: Qwen/Qwen3-4B
Training recipe: batched injector -> assessor self-play
Export path: outputs/local_training/medrect_mixed_v2_merged

This artifact is intended for evaluation and manual testing.

Downloads last month: 13

Safetensors

Model size

4B params

Tensor type

BF16

·

Video Preview

Reinforcement Learning

loading

Model tree for Abdine/qwen3-4b-medrect-mixed-r2

Base model

Qwen/Qwen3-4B-Base

Finetuned

Finetuned

(1158)

this model