Add author information and metadata

Hi! I'm Niels from the community science team at Hugging Face.

This PR improves the model card by:
- Adding the names of the authors who developed MERIT.
- Adding `base_model` and `datasets` metadata to link this repository to the MERT backbone and the training triplets.
- Adding the `library_name: pytorch` tag based on the provided sample code.
- Including a brief explanation of the three disentangled factors for better user understanding.

These changes help other researchers discover, attribute, and use your work more effectively!

README.md

@@ -1,7 +1,8 @@
 ---
-license: mit
 language:
 - en
+license: mit
+pipeline_tag: feature-extraction
 tags:
 - audio
 - music
@@ -10,19 +11,34 @@ tags:
 - contrastive-learning
 - music-information-retrieval
 - disentangled-representations
-pipeline_tag: feature-extraction
+library_name: pytorch
+base_model: m-a-p/MERT-v1-330M
+datasets:
+- amaai-lab/merit
 ---
 
 # MERIT — Disentangled Music Similarity Embeddings
 [![arXiv](https://img.shields.io/badge/arXiv-2605.27346-b31b1b.svg)](https://arxiv.org/abs/2605.27346)
 
-**MERIT** maps audio to three *disentangled* 128-dimensional unit vectors — one each for **melody**, **rhythm**, and **timbre** similarity. A single frozen [MERT-v1-330M](https://huggingface.co/m-a-p/MERT-v1-330M) backbone feeds three small trained projection heads that each specialize in one musical factor.
+**MERIT** (Multi-Factor Disentangled Music Similarity) maps audio to three *disentangled* 128-dimensional unit vectors — one each for **melody**, **rhythm**, and **timbre** similarity. A single frozen [MERT-v1-330M](https://huggingface.co/m-a-p/MERT-v1-330M) backbone feeds three small trained projection heads that each specialize in one musical factor.
+
+This model was presented in the paper [MERIT: Learning Disentangled Music Representations for Audio Similarity](https://huggingface.co/papers/2605.27346) by Abhinaba Roy, Junyi Liang, and [Dorien Herremans](https://huggingface.co/dorienh).
 
 > Code & training pipeline → [github.com/AMAAI-Lab/MERIT](https://github.com/AMAAI-Lab/MERIT)
 
-- [Paper](https://arxiv.org/abs/2605.27346)
+---
+
+## What is MERIT?
+
+Given two audio clips, MERIT returns **three independent cosine similarities** — one per musical factor:
+
+| Score | Captures | Example query |
+|---|---|---|
+| `S_mel` | Melodic contour & pitch identity | *"Find songs with the same melody"* |
+| `S_rhy` | Rhythmic groove & beat pattern | *"Find songs with the same drum feel"* |
+| `S_tim` | Instrument timbre & sonic character | *"Find songs played on the same instrument"* |
 
-- [Paper TLDR;](https://arxivexplained.com/papers/merit-learning-disentangled-music-representations-for-audio-similarity)
+A solo piano cover of a rock anthem scores high on `S_mel` but low on `S_rhy` and `S_tim`. MERIT makes this distinction explicit and computable.
 
 ---
 
@@ -178,4 +194,4 @@ If you use this model, please cite the paper in which it was presented:
 
 ## License
 
-[MIT](https://github.com/AMAAI-Lab/MERIT/blob/main/LICENSE)
+[MIT](https://github.com/AMAAI-Lab/MERIT/blob/main/LICENSE)