fix small typo

README.md

@@ -107,7 +107,7 @@ Following this, we perform instruction fine-tuning on [The Cauldron](https://hug
 - [atlas-math-sets](https://huggingface.co/datasets/AtlasUnified/atlas-math-sets)
 - [goat](https://huggingface.co/datasets/tiedong/goat)
 
-We use Lora to train the parameters initialized from pre-trained backbones and full fine-tuning for newly initialized parameters (modality connector), as we find this strategy to be more stable as long as more computationally efficient.
+We use Lora to train the parameters initialized from pre-trained backbones and full fine-tuning for newly initialized parameters (modality connector), as we find this strategy to be more stable as well as more computationally efficient.
 
 More details (training procedure, data selection, hyper-parameters, etc.) along with lessons learned from our ablations will be available in an upcoming technical report.