Self-ensemble: Mitigating Confidence Mis-calibration for Large Language Models
This addresses a specific reliability issue for users of LLMs in multi-choice tasks, but it is incremental as it builds on existing LLM architectures without fundamental changes.
The paper tackles the confidence distortion problem in Large Language Models (LLMs) on multi-choice question-answering, where models show under-confidence in correct predictions and over-confidence in incorrect ones, especially with many choices. It proposes Self-ensemble, a plug-and-play method that splits choices into groups and ensembles predictions, demonstrating improved performance over standard inference and baselines on three LLMs and datasets.
Although Large Language Models (LLMs) perform well in general fields, they exhibit a confidence distortion problem on multi-choice question-answering (MCQA), particularly as the number of answer choices increases. Specifically, on MCQA with many choices, LLMs suffer from under-confidence in correct predictions and over-confidence in incorrect ones, leading to a substantially degraded performance. To solve this problem, we propose Self-ensemble in this work. Our method splits the choices into several groups and ensembles LLM predictions across these groups to reach a final decision. The advantage of Self-ensemble is its plug-and-play nature, where it can be integrated into existing LLM architecture based on a designed attention mask and positional encoding, without requiring labeled datasets for parameter tuning. Experimental results on three LLMs and datasets demonstrate that Self-ensemble comprehensively addresses the confidence distortion problem of LLMs, outperforming standard inference as well as baseline methods.