TRL adds direct finetuning on agent action traces

Original: You can now finetune models on agent traces directly with TRL

Source: x.com ↗