Model checkpoints from the EMNLP 2025 paper "Improving Context Fidelity via Native Retrieval-Augmented Reasoning" by Wang et al.