Object-Level Verbalized Confidence Calibration in Vision-Language Models via Semantic Perturbation
This addresses the issue of user trust in VLMs by reducing misaligned confidence, though it is incremental as it builds on existing calibration methods.
The paper tackled the problem of poor calibration in vision-language models, where verbalized confidence often misaligns with response correctness, and proposed a Confidence Calibration through Semantic Perturbation (CSP) framework that significantly improved this alignment while maintaining task performance.
Vision-language models (VLMs) excel in various multimodal tasks but frequently suffer from poor calibration, resulting in misalignment between their verbalized confidence and response correctness. This miscalibration undermines user trust, especially when models confidently provide incorrect or fabricated information. In this work, we propose a novel Confidence Calibration through Semantic Perturbation (CSP) framework to improve the calibration of verbalized confidence for VLMs in response to object-centric queries. We first introduce a perturbed dataset where Gaussian noise is applied to the key object regions to simulate visual uncertainty at different confidence levels, establishing an explicit mapping between visual ambiguity and confidence levels. We further enhance calibration through a two-stage training process combining supervised fine-tuning on the perturbed dataset with subsequent preference optimization. Extensive experiments on popular benchmarks demonstrate that our method significantly improves the alignment between verbalized confidence and response correctness while maintaining or enhancing overall task performance. These results highlight the potential of semantic perturbation as a practical tool for improving the reliability and interpretability of VLMs.