Restore the statement that the Wan2.2 weights are fine-tuned.

ASSET_SHA256SUMS

@@ -1,8 +1,8 @@
 c485bd2ee62dbb1828eea5c5ea2bb7b0178f60d809329ebfd15bdf731f592f9a  action_vae.pt
 cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30  LICENSE
-93235892c29664be5451bec9c282ccf661bba6f6fc03a017bf6a00ec9d1a06d2  NOTICE
+b19048e39168d2101fa3330edc9574e09dc86b75bd3b7f6046f91b95ba508694  NOTICE
 b2cd77e9402b7897ba0d680eb45e4a38fb752e89641e3a9487a7cd4e947dad6c  policy.pt
-b460699373f8223f35d80b5da30863f225da0557bae4a92e9bfcf4ea46f44b56  README.md
+b4e53b1106e288caa42fd71d9ad8e8e8cf80c19b947442aef569be274dad6fe0  README.md
 835a23bad8ec61cf2c27c2f9440fe1f902ae5365c29f0c1858c0cb646e883594  stats/libero_10.json
 a648e545ad63d0d398ebf5735aa0e7426efdaf3fa55907c5ec50b72df6e273df  stats/libero_goal.json
 4eed60c361514c0f95f4ad24ec271972197e30463be07d82ec6c764cba3f6f83  stats/libero_object.json

NOTICE

@@ -1,13 +1,11 @@
 FastWAM LIBERO checkpoints
 Copyright (c) 2026 Advanced Micro Devices, Inc.
 
-These checkpoints include Wan2.2-TI2V-5B
+These checkpoints contain a modified version of Wan2.2-TI2V-5B
 (https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B, Wan Team, Alibaba), licensed
-under the Apache License, Version 2.0. The Wan2.2-TI2V-5B weights are the
-published weights and are unchanged. No data was used to train them
-separately. The action policy was trained with the FastWAM recipe
+under the Apache License, Version 2.0. The Wan2.2-TI2V-5B weights were
+fine-tuned by Advanced Micro Devices, Inc. using the FastWAM training recipe
 (https://github.com/yuantianyuan01/FastWAM, MIT License).
 
 Training data: LIBERO (Liu et al., NeurIPS 2023; CC BY 4.0), in the LeRobot
-format published as yuanty/LIBERO-fastwam (CC BY 4.0). The dataset is used as
-published. No changes were made to the training data.
+format published as yuanty/LIBERO-fastwam (CC BY 4.0).

README.md

@@ -12,7 +12,7 @@ tags:
 
 # FastWAM LIBERO
 
-A latent world-action policy for the [LIBERO](https://github.com/Lifelong-Robot-Learning/LIBERO) benchmark from Advanced Micro Devices, Inc. It uses the published [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) weights unchanged.
+A latent world-action policy for the [LIBERO](https://github.com/Lifelong-Robot-Learning/LIBERO) benchmark, fine-tuned by Advanced Micro Devices, Inc. from [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B).
 
 Each 32-step action chunk is encoded as 8 latent tokens of width 16. On the four LIBERO suites the policy reaches **97.2%** average success with a single sample.
 
@@ -44,7 +44,7 @@ huggingface-cli download Wan-AI/Wan2.1-T2V-1.3B \
   --local-dir checkpoints/Wan-AI/Wan2.1-T2V-1.3B
 ```
 
-The full Wan video DiT is not required. `policy.pt` already contains the published Wan2.2 video weights and the trained action expert.
+The full Wan video DiT is not required. `policy.pt` already contains the fine-tuned video and action experts.
 
 LIBERO simulator assets, BDDL task files, and initial states are also required.
 
@@ -62,9 +62,9 @@ Research on world-action models and robot-policy inference, in simulation.
 
 These weights are licensed under the Apache License 2.0 ([`LICENSE`](LICENSE)). See [`NOTICE`](NOTICE).
 
-The [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) weights included here are the published weights (Apache-2.0, Wan Team, Alibaba). They are unchanged, and no data was used to train them separately. The action policy was trained with the [FastWAM](https://github.com/yuantianyuan01/FastWAM) recipe (MIT). That license covers the training code, not these weights.
+They are a modified version of [Wan2.2-TI2V-5B](https://huggingface.co/Wan-AI/Wan2.2-TI2V-5B) (Apache-2.0, Wan Team, Alibaba), fine-tuned by Advanced Micro Devices, Inc. The training recipe is [FastWAM](https://github.com/yuantianyuan01/FastWAM) (MIT). That license covers the training code, not these weights.
 
-Training data: [LIBERO](https://github.com/Lifelong-Robot-Learning/LIBERO) (CC BY 4.0), in the LeRobot format published as [yuanty/LIBERO-fastwam](https://huggingface.co/datasets/yuanty/LIBERO-fastwam) (CC BY 4.0). This is the dataset used for FastWAM LIBERO, and it is used as published. No changes were made to the training data. The Wan2.2 weights are used unchanged. No data was used to train them separately.
+Training data: [LIBERO](https://github.com/Lifelong-Robot-Learning/LIBERO) (CC BY 4.0), in the LeRobot format published as [yuanty/LIBERO-fastwam](https://huggingface.co/datasets/yuanty/LIBERO-fastwam) (CC BY 4.0). This is the dataset used for FastWAM LIBERO, and it is used as published. No changes were made to the training data.
 
 ## Citation