Organize paper and updated checkpoints

README.md

@@ -11,14 +11,25 @@ tags:
 
 Official release checkpoints and benchmark artifacts for [OMG](https://github.com/Tsinghua-MARS-Lab/OMG).
 
-## Checkpoints
+## Paper Checkpoints
 
 | Model | Training step | File |
 |---|---:|---|
-| OMG 50M | 90,000 | `checkpoints/50m/sstep=090000.ckpt` |
-| OMG 100M | 100,000 | `checkpoints/100m/sstep=100000.ckpt` |
-| OMG 300M | 55,000 | `checkpoints/300m/sstep=055000.ckpt` |
-| OMG 500M | 50,000 | `checkpoints/500m/sstep=050000.ckpt` |
+| OMG 50M | 90,000 | `checkpoints/paper/50m/sstep=090000.ckpt` |
+| OMG 100M | 100,000 | `checkpoints/paper/100m/sstep=100000.ckpt` |
+| OMG 300M | 55,000 | `checkpoints/paper/300m/sstep=055000.ckpt` |
+| OMG 500M | 50,000 | `checkpoints/paper/500m/sstep=050000.ckpt` |
+
+## Updated Checkpoints
+
+The updated checkpoints were trained with diffusion loss only on the omni-modal LeRobot dataset.
+
+| Model | Training step | Status | File |
+|---|---:|---|---|
+| OMG 100M | 158,000 | Recommended | `updated/100m/sstep=158000.ckpt` |
+| OMG 300M | 154,000 | Experimental | `updated/300m/sstep=154000.ckpt` |
+
+The updated models improve the principal fixed-manifest text, audio-distribution, and HumanRef fidelity metrics over the paper 100M checkpoint. The 300M checkpoint remains experimental because simple-command rollout quality is not uniformly better than the updated 100M checkpoint.
 
 The pretrained benchmark evaluator is stored at `evaluator/step_004000.pt`.
 
@@ -29,4 +40,3 @@ The pretrained benchmark evaluator is stored at `evaluator/step_004000.pt`.
 - `SHA256SUMS`: checksums for release checkpoints and the evaluator.
 
 See the [OMG benchmark documentation](https://github.com/Tsinghua-MARS-Lab/OMG/blob/main/docs/benchmark.md) for evaluation commands and [OMG-Data](https://huggingface.co/datasets/THU-MARS/OMG-Data) for training data.
-

SHA256SUMS

@@ -1,5 +1,7 @@
-5e25dc7fd7251ec6df9dda8d3e008ff2540a5c8d1d691c576d4c64f9916eedaa  checkpoints/100m/sstep=100000.ckpt
-df4ce9aeced84fd96a39583c7f6e2c5773b76f942cb43012c6c51dd9eb806806  checkpoints/300m/sstep=055000.ckpt
-f13bc58f07f104fcff67e35e79650def0fec5a6eb62a193fa1ae9f979fb911cf  checkpoints/500m/sstep=050000.ckpt
-389fae1aa2ba31234e0ef621706362a19387ac485dce1f929408e06a3b012b2c  checkpoints/50m/sstep=090000.ckpt
+5e25dc7fd7251ec6df9dda8d3e008ff2540a5c8d1d691c576d4c64f9916eedaa  checkpoints/paper/100m/sstep=100000.ckpt
+df4ce9aeced84fd96a39583c7f6e2c5773b76f942cb43012c6c51dd9eb806806  checkpoints/paper/300m/sstep=055000.ckpt
+f13bc58f07f104fcff67e35e79650def0fec5a6eb62a193fa1ae9f979fb911cf  checkpoints/paper/500m/sstep=050000.ckpt
+389fae1aa2ba31234e0ef621706362a19387ac485dce1f929408e06a3b012b2c  checkpoints/paper/50m/sstep=090000.ckpt
+d45c93fb749bfe9d6e2875c60d40a6762c4777dca4f4f63eb59c6f1cdf106f52  updated/100m/sstep=158000.ckpt
+6c2520b76be7a4c76f7d2231d9832cb2c7b4130eb2e271115a52ce77b497ff7c  updated/300m/sstep=154000.ckpt
 e346790c5ce33c42840aa19f7fcc2891a4cab27e0d92e23bfaff7d48cb339c40  evaluator/step_004000.pt