OpenAI · AI & Cloud · Proposed by @MLEngineer

Fine-tuning with RLHF for custom models

Current fine-tuning only supports supervised learning. Allow customers to use RLHF with custom reward models to truly align fine-tuned models with specific business requirements.