Calibrate Before Use: Improving Few-Shot Performance of Language Models
This addresses the problem of unreliable few-shot performance for users of large language models, offering a practical solution to enhance stability and accuracy, though it is incremental as it builds on existing calibration techniques.
The authors tackled the instability in few-shot learning with language models like GPT-3, showing that accuracy varies widely based on prompt details, and they introduced a calibration method that improved average accuracy by up to 30.0% and reduced variance across tasks.
GPT-3 can perform numerous tasks when provided a natural language prompt that contains a few training examples. We show that this type of few-shot learning can be unstable: the choice of prompt format, training examples, and even the order of the training examples can cause accuracy to vary from near chance to near state-of-the-art. We demonstrate that this instability arises from the bias of language models towards predicting certain answers, e.g., those that are placed near the end of the prompt or are common in the pre-training data. To mitigate this, we first estimate the model's bias towards each answer by asking for its prediction when given the training prompt and a content-free test input such as "N/A". We then fit calibration parameters that cause the prediction for this input to be uniform across answers. On a diverse set of tasks, this contextual calibration procedure substantially improves GPT-3 and GPT-2's average accuracy (up to 30.0% absolute) and reduces variance across different choices of the prompt.