Add dataset card for EgoTeam

README.md

@@ -0,0 +1,46 @@
+---
+task_categories:
+- video-text-to-text
+tags:
+- robotics
+- multi-robot
+- egocentric-vision
+- spatial-reasoning
+---
+
+# EgoTeam: Multi-Robot Egocentric QA Dataset
+
+[**Paper**](https://huggingface.co/papers/2605.18431) | [**Code**](https://github.com/KPeng9510/seeing-together)
+
+**EgoTeam** is a large-scale multi-robot egocentric video question-answering dataset designed for cooperative embodied reasoning. It is part of the **CoopSR** benchmark, which evaluates the ability of Multimodal Large Language Models (MLLMs) to integrate synchronized egocentric videos from a team of moving robots to answer spatial, temporal, and coordination questions.
+
+## Dataset Highlights
+
+- **114K+ QA pairs**: Spanning 19 question types across four reasoning tiers.
+- **Team-based Reasoning**: Synchronized egocentric RGB-D videos from teams of 2, 3, and 4 robots.
+- **Simulated & Real-world**: Data collected in Habitat and iGibson simulators, with a real-world test set featuring quadruped robots.
+- **Rich Metadata**: Includes robot poses, pairwise relative poses, semantic information, object relations, and robot-object interactions.
+
+## Benchmark: CoopSR
+
+The CoopSR benchmark evaluates models across four progressive reasoning levels:
+
+| Tier | Name | Description |
+|---|---|---|
+| **T1** | Egocentric Spatial QA | Single-robot spatial awareness (locations, directions, layouts). |
+| **T2** | Pairwise Relationship Reasoning | Reasoning about two robots/viewpoints (visibility, occlusion, relations). |
+| **T3** | Scene-Level Composition | Integrating multiple views into a coherent scene-level representation and temporal movement. |
+| **T4** | Multi-Robot Dynamic Spatial Reasoning | High-level collaborative reasoning (team belief updates, coordination, task assignment). |
+
+## Citation
+
+If you find this dataset or benchmark useful for your research, please cite:
+
+```bibtex
+@article{peng2024seeing,
+  title={Seeing Together: Multi-Robot Cooperative Egocentric Spatial Reasoning with Multimodal Large Language Models},
+  author={Peng, Kun and others},
+  journal={arXiv preprint arXiv:2605.18431},
+  year={2024}
+}
+```