Add organization card with logo, model + dataset index, citation

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+logo.gif filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,10 +1,87 @@
 ---
-title: README
-emoji: 🦀
-colorFrom: purple
-colorTo: yellow
+title: TIDE-dllm
+emoji: 🌊
+colorFrom: blue
+colorTo: indigo
 sdk: static
 pinned: false
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+<center>
+<img src="logo.gif" width="320" />
+</center>
+
+<h1 align="center">Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models</h1>
+
+<p align="center">
+  🌊 The first cross-architecture distillation framework for diffusion LLMs — distilling 8B dense and 16B MoE teachers into a 0.6B student 🌊
+</p>
+
+<p align="center">
+  <a href="https://arxiv.org/abs/2604.26951">📄 arXiv 2604.26951</a> &nbsp;·&nbsp;
+  <a href="https://github.com/PKU-YuanGroup/TIDE">💻 Code (PKU-YuanGroup/TIDE)</a> &nbsp;·&nbsp;
+  <a href="https://pku-yuangroup.github.io/TIDE-Page/">🌐 Project page</a>
+</p>
+
+---
+
+This organization hosts the **distilled student checkpoints** and **pre-tokenized SFT datasets** released with TIDE. The framework consists of three modular components — **TIDAL** (dual-axis interpolation), **CompDemo** (complementary mask-split teacher inference), and **Reverse CALM** (cross-tokenizer chunk-level matching) — and is evaluated across two heterogeneous distillation pipelines.
+
+## ✨ Highlights
+
+- **+1.53 average gain** over the non-distilled BD3LM baseline across 8 benchmarks (34.20 vs. 32.67).
+- **+16.48 on HumanEval** over the equivalent-size AR baseline (48.78 vs. 32.30) — distilled dLLMs especially excel at code generation.
+- **22× peak-memory reduction** vs. the 16B MoE LLaDA2 teacher (1.4 GB vs. 31.3 GB) and **5.2× faster inference** (6.25 s vs. 32.55 s for 256 tokens on H100), enabling commodity-hardware deployment.
+
+## 🤖 Released models
+
+Six 0.6B distilled student checkpoints (3 per pipeline). Each is initialized from [`dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) and distilled from a larger dLLM teacher.
+
+| Pipeline | Variant | Repo |
+|---|---|---|
+| A — Cross-Tokenizer (LLaDA2 teacher) | **TIDE-Cross** (native, paper-best) | [distill-LLaDA2-TIDE_Cross](https://huggingface.co/TIDE-dllm/distill-LLaDA2-TIDE_Cross) |
+| A — Cross-Tokenizer (LLaDA2 teacher) | TIDE-Shared variant | [distill-LLaDA2-TIDE_Shared](https://huggingface.co/TIDE-dllm/distill-LLaDA2-TIDE_Shared) |
+| A — Cross-Tokenizer (LLaDA2 teacher) | CALM baseline | [distill-LLaDA2-CALM](https://huggingface.co/TIDE-dllm/distill-LLaDA2-CALM) |
+| B — Shared-Tokenizer (WeDLM teacher) | **TIDE-Shared** (native, paper-best) | [distill-WeDLM-TIDE_Shared](https://huggingface.co/TIDE-dllm/distill-WeDLM-TIDE_Shared) |
+| B — Shared-Tokenizer (WeDLM teacher) | TIDE-Cross variant | [distill-WeDLM-TIDE_Cross](https://huggingface.co/TIDE-dllm/distill-WeDLM-TIDE_Cross) |
+| B — Shared-Tokenizer (WeDLM teacher) | KL baseline | [distill-WeDLM-KL](https://huggingface.co/TIDE-dllm/distill-WeDLM-KL) |
+
+## 📚 Released datasets
+
+Pre-tokenized SFT mixtures (`tulu-3-sft-mixture` + `smoltalk` + `opc-sft-stage1` + `opc-sft-stage2`) prepared for each teacher, so distillation jobs never have to re-tokenize at startup.
+
+| Pipeline | Repo |
+|---|---|
+| A — for the LLaDA2 teacher | [distill_llada2_sft](https://huggingface.co/datasets/TIDE-dllm/distill_llada2_sft) |
+| B — for the WeDLM teacher | [distill_wedlm_sft](https://huggingface.co/datasets/TIDE-dllm/distill_wedlm_sft) |
+
+## 🚀 Quick start
+
+```python
+import torch
+from transformers import AutoModelForMaskedLM, AutoTokenizer
+
+repo = "TIDE-dllm/distill-LLaDA2-TIDE_Cross"   # paper-best Pipeline-A checkpoint
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+model = AutoModelForMaskedLM.from_pretrained(
+    repo, dtype=torch.bfloat16, trust_remote_code=True,
+).to(device).eval()
+tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
+```
+
+The same `generate()` routine published with [`dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) works on every TIDE checkpoint — just swap the model name.
+
+## 📝 Citation
+
+```bibtex
+@misc{zhang2026turningtidecrossarchitecturedistillation,
+      title={Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models},
+      author={Gongbo Zhang and Wen Wang and Ye Tian and Li Yuan},
+      year={2026},
+      eprint={2604.26951},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2604.26951},
+}
+```