Fix minor issues

4D-R1-GRPO-MP/README.md

@@ -6,4 +6,4 @@ Use this adapter with the VerseCrafter base model and the 4D-R1 inference implem
 
 LoRA rank: 16. Target modules: q, k, v, o.
 
-MP denotes mixed-precision training; it does not mean that the saved adapter is an FP4-quantized standalone model.
+MP denotes mixed-precision training.

4D-R1-NFT-MP/README.md

@@ -6,4 +6,4 @@ Use this adapter with the VerseCrafter base model and the 4D-R1 inference implem
 
 LoRA rank: 16. Target modules: q, k, v, o.
 
-MP denotes mixed-precision training; it does not mean that the saved adapter is an FP4-quantized standalone model.
+MP denotes mixed-precision training.

README.md

@@ -21,8 +21,8 @@ Each directory contains `adapter_model.safetensors` and `adapter_config.json`. T
 
 These are LoRA adapters, not standalone full-model checkpoints. Inference requires the VerseCrafter base model, its companion components, and the compatible 4D-R1 inference implementation. Base-model weights are not included in this repository.
 
-MP denotes mixed-precision training. It does not indicate that the saved adapter weights form an FP4-quantized model.
+MP denotes mixed-precision training.
 
-The corresponding paper rows are GRPO, NFT, GRPO (mixed-precision), and NFT (mixed-precision). Internal run identifiers and update counts are not part of the public checkpoint names.
+The corresponding paper rows are GRPO, NFT, GRPO (mixed-precision), and NFT (mixed-precision).
 
 `SHA256SUMS` provides checksums for the adapter weights and configuration files.