Add files using upload-large-folder tool

README.md

@@ -1,23 +1,70 @@
-# MM_eval Packaged Dataset
+<div align="center">
 
-This directory contains tar shards prepared for upload to the Hugging Face dataset repo `xhLiu/MM_eval`.
+<h1><a color="red" href="https://arxiv.org/pdf/2511.12034">Calibrated Multimodal Representation Learning with Missing Modalities</a></h1>
 
-Packaging choices:
-- Each top-level dataset under `datasets/` is archived independently.
-- Archives are split into 5 GiB shards named `DATASET.tar.part-XXX`.
-- Local cache directories such as `audiocaps_train/.cache` are excluded.
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
+![License](https://img.shields.io/badge/Accepted-ICML'2026-red)
 
-To restore a dataset after download:
+*Multimodal representation learning under partial-modality settings*
 
-```bash
-cat DATASET.tar.part-* | tar -xf -
-```
+</div>
+
+## ✨ Overview
+
+<p align="center">
+  <img src="img/anchor_shift.jpg" alt="Anchor shift" width="420" />
+</p>
+
+**CalMRL** is a multimodal representation learning framework designed for alignment calibration when some modalities are missing.
+CalMRL combines two complementary goals:
+- **Cross-modal alignment** for robust shared representations
+- **Missing-modality calibration** through posterior inference and learned generative parameters
+
+---
+
+## 🎯 Key Features
+
+🔄 **Partial-Modality Learning**
+- Handles missing video, audio, text, or subtitle signals
+- Supports posterior-based feature completion with learned modality-specific parameters
+
+🎯 **Multimodal Retrieval**
+- Joint training over text-video, text-audio, text-video-audio, and subtitle-aware setups
+- Config-driven recipes for pretraining, finetuning, and evaluation
+
+🧠 **Feature Calibration**
+- Uses latent posterior inference for modality completion
+- Includes a warmup pipeline to estimate `W`, `mu`, and `log_sigma`
 
-Examples:
+---
 
-```bash
-cat activitynet.tar.part-* | tar -xf -
-cat vatex.tar.part-* | tar -xf -
+## 🏗️ Architecture
+
+![](img/framework.png)
+
+The current codebase is organized around three main stages:
+
+1. **🔧 Multimodal Encoding**: Video, audio, text, and subtitle features are extracted with VAST-style encoders.
+2. **🧮 Representation Calibration**: Shared embeddings are aligned while latent posterior inference estimates missing information.
+3. **🔄 Downstream Evaluation**: Retrieval and other tasks are executed through a unified config-driven pipeline.
+
+---
+
+## Citation
+
+If this project is useful for your research, you can cite the work as:
+
+```bibtex
+@article{liu2025calibrated,
+  title={Calibrated Multimodal Representation Learning with Missing Modalities},
+  author={Liu, Xiaohao and Xia, Xiaobo and Wei, Jiaheng and Yang, Shuo and Su, Xiu and Ng, See-Kiong and Chua, Tat-Seng},
+  journal={arXiv preprint arXiv:2511.12034},
+  year={2025}
+}
 ```
 
-Checksums for all shards are stored in `sha256sums.txt`.
+<div align="center">
+
+**[🔝 Back to Top](#-overview)**
+
+</div>