Improve model card and add metadata

README.md

@@ -1,8 +1,23 @@
 ---
 license: apache-2.0
+pipeline_tag: robotics
 ---
 
-[Arxiv Paper](https://arxiv.org/abs/2603.04639) |  [HF Paper](https://huggingface.co/papers/2603.04639) |  [Website](https://robomme.github.io/) | [Benchmark Code](https://github.com/RoboMME/robomme_benchmark) | [Policy Learning Code](https://github.com/RoboMME/robomme_policy_learning)
+# RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies
 
+[Arxiv Paper](https://arxiv.org/abs/2603.04639) | [HF Paper](https://huggingface.co/papers/2603.04639) | [Website](https://robomme.github.io/) | [Benchmark Code](https://github.com/RoboMME/robomme_benchmark) | [Policy Learning Code](https://github.com/RoboMME/robomme_policy_learning)
 
-Single model ckpt for perceptual-framesamp-modul 80k.
+This repository contains the model checkpoint for `perceptual-framesamp-modul` trained for 80k steps. This model is part of the **RoboMME** suite, a large-scale standardized benchmark for evaluating and advancing Vision-Language-Action (VLA) models in long-horizon, history-dependent robotic manipulation scenarios.
+
+RoboMME introduces a taxonomy evaluating temporal, spatial, object, and procedural memory across 16 manipulation tasks. This specific checkpoint is one of the 14 memory-augmented VLA variants built on the $\pi_{0.5}$ backbone explored in the research.
+
+## Citation
+
+```bibtex
+@article{dai2026robomme,
+  title={RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies},
+  author={Dai, Yinpei and Fu, Hongze and Lee, Jayjun and Liu, Yuejiang and Zhang, Haoran and Yang, Jianing and Finn, Chelsea and Fazeli, Nima and Chai, Joyce},
+  journal={arXiv preprint arXiv:2603.04639},
+  year={2026}
+}
+```