De-anonymize and polish README: add ICML 2026 paper info, arXiv DOI, cross-links to vocsim/* repos

README.md

@@ -26,53 +26,82 @@ configs:
   data_files:
   - split: train
     path: data/train-*
+license: cc-by-4.0
+tags:
+- audio
+- animal-vocalization
+- ultrasonic-vocalization
+- mouse
+- bioacoustics
+- speaker-identification
+- benchmark
+- vocsim
+size_categories:
+- 10K<n<100K
+pretty_name: VocSim — Mouse Identity Classification
 ---
-# Dataset Card for VocSim - Mouse Identity Classification
 
-## Dataset Description
+# VocSim — Mouse Identity Classification
 
-This dataset is used in the **VocSim benchmark** paper for evaluating the ability of neural audio embeddings to identify individual mice based on their ultrasonic vocalization (USV) syllables. It contains pre-segmented USV syllables from multiple individual mice, derived from recordings created by **Van Segbroeck et al. (2017)**.
+[![GitHub](https://img.shields.io/badge/GitHub-vocsim%2Fbenchmark-black?logo=github)](https://github.com/vocsim/benchmark)
+[![Core dataset](https://img.shields.io/badge/%F0%9F%A4%97%20Core-vocsim%2Fpublic-blue)](https://huggingface.co/datasets/vocsim/public)
+[![License: CC BY 4.0](https://img.shields.io/badge/License-CC%20BY%204.0-blue.svg)](https://creativecommons.org/licenses/by/4.0/)
 
-The primary task associated with this dataset is supervised classification: training a model to predict the correct mouse identity (`speaker` field) given an audio input (a single syllable) or its derived features.
+A companion dataset for the **VocSim** benchmark that tests whether audio embeddings preserve **individual identity** in mouse ultrasonic vocalizations (USVs). It contains pre-segmented USV syllables from multiple individual mice (the `speaker` field), sampled at the native 250 kHz, derived from recordings by Van Segbroeck et al. (2017).
 
-**Included Files:**
-*   Hugging Face `Dataset` object containing audio file paths (individual syllables) and metadata (mouse identity).
+> Basha, M., Zai, A. T., Stoll, S., & Hahnloser, R. H. R. *VocSim: A Training-free Benchmark for Zero-shot Content Identity in Single-source Audio.* ICML 2026. [arXiv:2512.10120](https://doi.org/10.48550/arXiv.2512.10120)
 
-## Dataset Structure
+## Task
 
-### Data Instances
+Supervised multi-class classification: given an audio syllable (or features derived from it), predict which individual mouse produced it. The target is `speaker`. In the paper we use this dataset to validate that VocSim-top embeddings transfer to a fine-grained downstream bioacoustic task.
 
-A typical example in the dataset looks like this:
+## Schema
 
 ```python
 {
-  'audio': {'path': '/path/to/datasets/mouse_identity/BM003/BM003_syllable_1.wav', 'array': array([...], dtype=float32), 'sampling_rate': 250000},
-  'subset': 'mouse_identity',
-  'index': 50,
-  'speaker': 'BM003', # The crucial individual mouse ID (target label)
-  'label': 'BM003_syllable_1', # Syllable-specific identifier
-  'original_name': 'BM003/BM003_syllable_1.wav' # Example original relative path
+  "audio":  {"array": np.ndarray, "sampling_rate": 250000},
+  "subset": "mouse_identity",
+  "index":  50,
+  "speaker": "BM003",                  # target: individual mouse ID
+  "label":   "BM003_syllable_1",       # syllable-specific identifier
+  "original_name": "BM003/BM003_syllable_1.wav"
 }
 ```
-### Citation Information
 
-If you use this dataset, please cite the VocSim benchmark paper, and the MUPET software if relying on the provided segmentation:
+## Quick start
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("vocsim/mouse-identity-classification-benchmark", split="train")
+print(ds[0])
 ```
-@unpublished{vocsim2025,
-  title={VocSim: A Training-Free Benchmark for Content Identity in Single-Source Audio Embeddings},
-  author={Anonymous},
-  year={2025},
-  note={Submitted manuscript}
+
+For end-to-end evaluation, use [github.com/vocsim/benchmark](https://github.com/vocsim/benchmark) — see `reproducibility/scripts/mouse_identity.py`.
+
+## Source data
+
+USV recordings and segmentation rely on MUPET (Van Segbroeck et al., 2017). Please cite both that work and the VocSim paper if you use this dataset.
+
+## Citation
+
+```bibtex
+@inproceedings{basha2026vocsim,
+  title     = {VocSim: A Training-free Benchmark for Zero-shot Content Identity in Single-source Audio},
+  author    = {Basha, Maris and Zai, Anja T. and Stoll, Sabine and Hahnloser, Richard H. R.},
+  booktitle = {Proceedings of the 43rd International Conference on Machine Learning (ICML)},
+  year      = {2026},
+  doi       = {10.48550/arXiv.2512.10120}
 }
 
 @article{VanSegbroeck2017,
-    author = {Van Segbroeck, Maarten and Knoll, Aaron T. and Levitt, Patricia and Narayanan, Shrikanth},
-    title = "{MUPET}-Mouse Ultrasonic Profile ExTraction: A Signal Processing Tool for Rapid and Unsupervised Analysis of Ultrasonic Vocalizations",
-    journal = {Neuron},
-    volume = {94},
-    number = {3},
-    pages = {465--485.e5},
-    year = {2017},
-    doi = {10.1016/j.neuron.2017.04.018}
+  author  = {Van Segbroeck, Maarten and Knoll, Aaron T. and Levitt, Patricia and Narayanan, Shrikanth},
+  title   = {{MUPET}-Mouse Ultrasonic Profile ExTraction: A Signal Processing Tool for Rapid and Unsupervised Analysis of Ultrasonic Vocalizations},
+  journal = {Neuron},
+  volume  = {94},
+  number  = {3},
+  pages   = {465--485.e5},
+  year    = {2017},
+  doi     = {10.1016/j.neuron.2017.04.018}
 }
-```
+```