FESTA: Functionally Equivalent Sampling for Trust Assessment of Multimodal LLMs

2025-09-20

AI Summary

Assessing the trustworthiness of multimodal large language models (MLLMs) is challenging due to input heterogeneity. This paper proposes FESTA, a black-box trust assessment method that introduces, for the first time, task-preserving equivalent and complementary input sampling—enabling uncertainty estimation via functional equivalence and complementarity modeling without requiring ground-truth labels or internal model access. FESTA extends the multimodal input space solely through the input-output interface, achieving unsupervised, plug-and-play trust quantification. Evaluated on visual and audio reasoning tasks, FESTA improves AUROC for misprediction detection by 33.3% and 29.6%, respectively, significantly enhancing selective prediction capability and user trust. The implementation is publicly available.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

The accurate trust assessment of multimodal large language models (MLLMs) generated predictions, which can enable selective prediction and improve user confidence, is challenging due to the diverse multi-modal input paradigms. We propose Functionally Equivalent Sampling for Trust Assessment (FESTA), a multimodal input sampling technique for MLLMs, that generates an uncertainty measure based on the equivalent and complementary input samplings. The proposed task-preserving sampling approach for uncertainty quantification expands the input space to probe the consistency (through equivalent samples) and sensitivity (through complementary samples) of the model. FESTA uses only input-output access of the model (black-box), and does not require ground truth (unsupervised). The experiments are conducted with various off-the-shelf multi-modal LLMs, on both visual and audio reasoning tasks. The proposed FESTA uncertainty estimate achieves significant improvement (33.3% relative improvement for vision-LLMs and 29.6% relative improvement for audio-LLMs) in selective prediction performance, based on area-under-receiver-operating-characteristic curve (AUROC) metric in detecting mispredictions. The code implementation is open-sourced.

Problem

• Assessing trust in multimodal LLM predictions without ground truth
• Quantifying uncertainty for diverse multimodal input paradigms
• Enabling selective prediction through black-box model evaluation

Innovation

• Functionally Equivalent Sampling for uncertainty quantification
• Black-box unsupervised trust assessment technique
• Expands input space with equivalent and complementary samples