Add links to TRL docs

app/src/content/article.mdx

@@ -76,7 +76,7 @@ Building on Universal Logit Distillation (ULD) [@boizard2025crosstokenizerdistil
 
 Our key contributions are:
 
-- Providing an open-source implementation of on-policy distillation methods and proving they work for multiple model combinations.
+- Providing an open-source implementation of on-policy distillation methods in TRL ([GKD](https://huggingface.co/docs/trl/en/gkd_trainer) and [GOLD](https://huggingface.co/docs/trl/main/en/gold_trainer)) and proving they work for multiple model combinations.
 - Extending ULD to the on-policy setting, where we sample completions from the student and align them to the teacher's distribution.
 - Implementing new sequence and vocabulary alignment methods that improve distillation performance when the student and the teacher have different tokenizers.