Update model card

README.md

@@ -1,43 +1,78 @@
 ---
 license: apache-2.0
 library_name: pytorch
+datasets:
+  - Gabriel166/MORAD-data
 tags:
-- rna
-- inverse-folding
-- reinforcement-learning
-- diffusion
-- grpo
+  - rna
+  - rna-design
+  - inverse-folding
+  - diffusion
+  - reinforcement-learning
 ---
-# RIDE-RL v6f step 390
 
-This repository contains the selected RIDE-RL post-training checkpoint used in the 153-target evaluation.
+# MORAD
 
-## Files
+Checkpoint for **MORAD: Reinforcement Learning with Multi-Objective Rewards for
+RNA Inverse Design**: the pretrained RIDE backbone-conditioned RNA diffusion
+model after 390 updates of online reinforcement learning with the six-component
+MORAD reward. A single shared policy designs nucleotide sequences for any target
+3D backbone, with no per-target optimization.
 
-- `policy_step000390.pt`: PyTorch checkpoint with a single top-level `model` state dictionary.
-- `config/rl_v6f_t75.yaml`: exact long-training configuration used for run v6f.
-- `SHA256SUMS`: file checksums.
+Code: [github.com/Gabrile166/MORAD](https://github.com/Gabrile166/MORAD) ·
+Data: [Gabriel166/MORAD-data](https://huggingface.co/datasets/Gabriel166/MORAD-data)
 
-## Model and training
+## File
 
-The policy starts from the pretrained RIDE backbone and is post-trained with grouped policy updates and a geometric structural reward. The v6f run used 527 targets, 3 epochs, group size 8, 4 targets per update, learning rate `2e-4`, and 395 optimizer updates. Validation was run every 10 updates on 153 targets.
+| File | Content |
+|---|---|
+| `policy_step000390.pt` | MORAD policy after 390 updates, `{"model": state_dict}` (10,214,213 parameters) |
 
-The selected snapshot is step 390 (the last scheduled validation snapshot; training ended at step 395). On the 153-target, candidate-0 evaluation it achieved GDT-TS 0.3827, C4'-TM 0.3400, RMSD 7.9196 Å, C1'-TM 0.3855 (n=126), success rate 0.3007, and sequence recovery 0.5056. See the project repository for evaluation code and metric definitions.
+SHA-256 checksums are listed in `SHA256SUMS`.
 
-## Loading
+## Results on the 153 test targets
 
-```python
-import torch
-checkpoint = torch.load("policy_step000390.pt", map_location="cpu")
-state_dict = checkpoint["model"]
+Structural and pairing metrics use candidate 0 of each target (structures
+predicted with RhoFold+, pairs with EternaFold); NED is from the thermodynamic
+evaluation (ViennaRNA).
+
+| Model | GDT-TS ↑ | TM (C4′) ↑ | RMSD (Å) ↓ | TM (C1′) ↑ | MCC ↑ | NED ↓ |
+|---|---:|---:|---:|---:|---:|---:|
+| RiboDiffusion | 0.3132 | 0.2839 | 10.9215 | 0.3613 | 0.3267 | 0.6056 |
+| gRNAde (T=0.1) | 0.3591 | 0.3179 | 8.8374 | 0.3642 | 0.6734 | 0.3458 |
+| RDesign | 0.3291 | 0.2902 | 10.3441 | 0.3392 | 0.6566 | 0.3349 |
+| RIDE | 0.3300 | 0.2841 | 9.7199 | 0.3538 | 0.6110 | 0.3906 |
+| **MORAD** | **0.3827** | **0.3400** | **7.9196** | **0.3855** | **0.7295** | **0.2830** |
+
+All tables of the paper are in the
+[repository README](https://github.com/Gabrile166/MORAD#results).
+
+## Usage
+
+```bash
+git clone https://github.com/Gabrile166/MORAD.git && cd MORAD
+# install as described in docs/INSTALL.md, then
+bash scripts/download_assets.sh            # model weights, data and this checkpoint
+bash scripts/evaluate_test153.sh morad     # full evaluation on the 153 test targets
 ```
 
-The architecture and runtime are provided by the RIDE-RL code repository: https://github.com/Gabrile166/RIDE-RL
+`download_assets.sh` stores this file as `artifacts/models/morad/morad_update390.pt`.
+Any MORAD checkpoint can be evaluated with
+`RIDE_CHECKPOINT=path/to/policy.pt python evaluate.py --config configs/eval/test153.yaml`.
 
-## Dependencies and license
+```python
+import torch
 
-This checkpoint is a derivative of the Apache-2.0 RIDER/RIDE implementation and is released under Apache-2.0. It does not include RhoFold+, USalign, EternaFold, or other third-party weights/binaries; obtain those from their original sources under their own terms.
+checkpoint = torch.load("policy_step000390.pt", map_location="cpu")
+state_dict = checkpoint["model"]  # RIDE network, defined in configs/ride_policy.yaml
+```
 
 ## Citation
 
-Please cite the upstream RIDER work and this repository when using the checkpoint.
+```bibtex
+@misc{tang2026morad,
+  title  = {{MORAD}: Reinforcement Learning with Multi-Objective Rewards for {RNA} Inverse Design},
+  author = {Tang, Jixin and Guo, Ji and Zhao, Jun},
+  year   = {2026}
+}
+```