Which training method aligns language models with human feedback?

Which training method aligns language models with human feedback?

• A. API
• B. Dataset
• C. RLHF
• D. Open-Source Model

Answer: (C)

Reinforcement Learning from Human Feedback (RLHF) is the training approach that directly uses human judgments to shape what the model should prefer in its outputs. In practice, humans review multiple possible responses to prompts and rank or score them. A reward model learns to predict these human preferences, and the language model is then fine-tuned with reinforcement learning to maximize that reward. The result is a model whose outputs align more closely with what people consider helpful, safe, and accurate. Other options aren’t training methods that embed human preferences. An API is just a way to interact with the model, not a training technique. A dataset is the material used for training but doesn’t inherently optimize for human judgments. An open-source model describes accessibility and licensing, not the alignment method. RLHF uniquely uses human feedback to guide the training process toward desirable behavior.

Reinforcement Learning from Human Feedback (RLHF) is the training approach that directly uses human judgments to shape what the model should prefer in its outputs. In practice, humans review multiple possible responses to prompts and rank or score them. A reward model learns to predict these human preferences, and the language model is then fine-tuned with reinforcement learning to maximize that reward. The result is a model whose outputs align more closely with what people consider helpful, safe, and accurate.

Other options aren’t training methods that embed human preferences. An API is just a way to interact with the model, not a training technique. A dataset is the material used for training but doesn’t inherently optimize for human judgments. An open-source model describes accessibility and licensing, not the alignment method. RLHF uniquely uses human feedback to guide the training process toward desirable behavior.