Instruction Following
The ability of an AI model to understand and execute tasks according to instructions provided in user or system prompts. It is a foundational metric for measuring model controllability and alignment, and is typically achieved through Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF).
Timeline (last 90 days)
Instruction following refers to a model accurately understanding and executing the constraints given by users in natural language, including format, scope and negative instructions.
In the context of large language models, 'intent understanding' generally refers to the model's ability to map a user's natural language instructions or behavioral patterns to the user's actual underlying goals, a topic that academia sometimes places under the research scope of 'instruction following' or 'task alignment'.
This method measures answer quality along three core evaluation axes: Composition, Grounding, and Instruction-Following
All Facts (3)
Instruction following refers to a model accurately understanding and executing the constraints given by users in natural language, including format, scope and negative instructions.
50%UnverifiedIn the context of large language models, 'intent understanding' generally refers to the model's ability to map a user's natural language instructions or behavioral patterns to the user's actual underlying goals, a topic that academia sometimes places under the research scope of 'instruction following' or 'task alignment'.
50%UnverifiedThis method measures answer quality along three core evaluation axes: Composition, Grounding, and Instruction-Following
50%