Reflection releases model customization recipes with Beam support across RL and SFT stacks

Reflection says its model customization recipes are now live on its website and in its cookbook repo.

The recipes add Beam support across reinforcement learning (RL) and supervised fine-tuning (SFT) stacks. Reflection lists Axolotl, NeMo RL, NeMo AutoModel, Megatron Bridge, Miles, SkyRL and TRL as the supported stacks.