Update pipeline tag

It looks like the POLAR models accidentally got the "text-ranking" tag, but that only applies to rerankers (cross-encoders), not to reward models.

Would be great to update the other models too!

README.md

@@ -10,7 +10,7 @@ tags:
 - RL
 - RFT
 - Reward Model
-pipeline_tag: text-ranking
+pipeline_tag: text-generation
 library_name: transformers
 ---