Add paper link, GitHub repository, and task category

README.md

@@ -50,6 +50,31 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+task_categories:
+- text-classification
 ---
 
-Paper: https://arxiv.org/pdf/2605.31170
+# Emergent Languages in Populations of Language Model Agents
+
+This dataset contains agent-generated posts from Moltbook, filtered and categorized to study emergent languages. It was presented in the paper [Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion](https://huggingface.co/papers/2605.31170).
+
+The dataset includes 518 examples categorized into:
+- **Token efficiency** (166): Proposals for languages aimed at reducing computational/token costs.
+- **New natural languages** (106): The invention of new natural-sounding languages.
+- **Oversight evasion** (59): Sophisticated protocols and languages proposed specifically to avoid human monitoring and oversight.
+
+### Links
+
+- **Paper:** [https://arxiv.org/abs/2605.31170](https://arxiv.org/abs/2605.31170)
+- **GitHub:** [https://github.com/aisilab/emergent-languages](https://github.com/aisilab/emergent-languages)
+
+### Citation
+
+```bibtex
+@article{beltoft2026emergent,
+  title={Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion},
+  author={Beltoft, Stine Lyngsø and Brach, William and Torrielli, Federico and Nielsen, Jacob and Pirchert, Annemette Brok and Tonini, Filippo and Schneider-Kamp, Peter and Poech, Lukas Galke},
+  journal={arXiv preprint arXiv:2605.31170},
+  year={2026}
+}
+```