ReflectRL is a lightweight framework for learning from Golden Negative Trajectories (GNTs) during on-policy post-training. Instead of imitating failed expert trajectories directly, ReflectRL uses them as reflective context during training and gradually transitions the policy back to direct reasoning for inference.

Citation

Code &amp; Repository: <a href="https://github.com/bibisbar/ReflectRL" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">GitHub - ReflectRL
Base Model: <a href="https://huggingface.co/Qwen/Qwen2.5-Math-7B" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">Qwen/Qwen2.5-Math-7B
Paper: <a href="https://huggingface.co/papers/2608.03972" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning

bibtex

@article{reflectrl2027,
  title={ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning},
  author={Jinhe Bi and Chennan Zhou and Zengjie Jin and Aniri and Shuo Lu and Wenke Huang and Hu Cao and Xun Xiao and Zhihong Zhu and Volker Tresp and Fei Shen and Yunpu Ma and Tat-Seng Chua},
  year={2027}
}

Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO-ReflectRL

ReflectRL

Overview

Citation

Jinhe/
ReflectRL-Qwen2.5-Math-7B-GRPO-ReflectRL