Vision-language models (VLMs) are increasingly used as automated judges for multimodal systems, yet their scores provide no indication of reliability. We study this problem through conformal prediction, a distribution-free framework that converts a judge's point score into a calibrated prediction interval using only score-token log-probabilities, with no retraining. We present the first systematic analysis of conformal prediction for VLM-as-a-Judge across 3 judges and 14 visual task categories. Our results show that evaluation uncertainty is strongly task-dependent: intervals cover ~40% of the score range for aesthetics and natural images but expand to ~70% for chart and mathematical reasoning, yielding a quantitative reliability map for multimodal evaluation. We further identify a failure mode not captured by standard evaluation metrics, ranking-scoring decoupling, where judges achieve high ranking correlation while producing wide, uninformative intervals, correctly ordering responses but failing to assign reliable absolute scores. Finally, we show that interval width is driven primarily by task difficulty and annotation quality, i.e., the same judge and method yield 4.5x narrower intervals on a clean, multi-annotator captioning benchmark. Code: https://github.com/divake/VLM-Judge-Uncertainty

Large Language Models (LLMs) are widely used as proxies for human labelers in both training (Reinforcement Learning from AI Feedback) and large-scale response evaluation (LLM-as-a-judge). Alignment and evaluation are critical components in the development of reliable LLMs, and the choice of feedback protocol plays a central role in both but remains understudied. In this work, we show that the choice of feedback protocol for evaluation (absolute scores versus relative preferences) can significantly affect evaluation reliability and induce systematic biases. In the context of LLM-as-a-judge evaluation, we show that pairwise protocols are more vulnerable to distracted evaluation. Generator models can exploit spurious attributes (or distractor features) favored by the LLM judge, resulting in inflated scores for lower-quality outputs. We find that absolute scoring is more robust to such manipulation, producing judgments that better reflect response quality and are less influenced by distractor features. Our results demonstrate that generator models can flip preferences by embedding distractor features, skewing LLM-as-a-judge comparisons and leading to inaccurate conclusions about model quality in benchmark evaluations. Pairwise preferences flip in about 35% of the cases, compared to only 9% for absolute scores. We offer recommendations for choosing feedback protocols based on dataset characteristics and evaluation objectives.

The expansion of the Internet of Things (IoT) in the battlefield, Internet of Battlefield Things (IoBT), gives rise to new opportunities for enhancing situational awareness. To increase the potential of IoBT for situational awareness in critical decision making, the data from these devices must be processed into consumer-ready information objects, and made available to consumers on demand. To address this challenge we propose a workflow that makes use of natural language processing (NLP) to query a database technology and return a response in natural language. Our solution utilizes Large Language Models (LLMs) that are sized for edge devices to perform NLP as well as graphical databases which are well suited for dynamic connected networks which are pervasive in the IoBT. Our architecture employs LLMs for both mapping questions in natural language to Cypher database queries as well as to summarize the database output back to the user in natural language. We evaluate several medium sized LLMs for both of these tasks on a database representing publicly available data from the US Army's Multipurpose Sensing Area (MSA) at the Jornada Range in Las Cruces, NM. We observe that Llama 3.1 (8 billion parameters) outperforms the other models across all the considered metrics. Most importantly, we note that, unlike current methods, our two step approach allows the relaxation of the Exact Match (EM) requirement of the produced Cypher queries with ground truth code and, in this way, it achieves a 19.4% increase in accuracy. Our workflow lays the ground work for deploying LLMs on edge devices to enable natural language interactions with databases containing information objects for critical decision making.

Generative models have achieved remarkable success across numerous applications, showcasing their versatility and effectiveness in domains such as image synthesis, natural language processing, and scientific discovery (Achiam et al. 2023; Goodfellow et al. 2014; Karras et al. 2020; Van Den Oord et al. 2016). While extensive research has focused on developing and refining generative models, comparatively less attention has been given to evaluating these models. Evaluating generative models is essential for quantifying the quality of their outputs and identifying the best model when comparing multiple options. Evaluating a generative model is significantly more challenging than the evaluation of a predictor or a classifier. To evaluate the performance of prediction or classification, we can directly compare the model's output with the true label. In contrast, the quality of a generative model is determined by how closely the distribution of its generated data matches that of the input data, rather than the similarity between generated data points and input data points (also known as the reconstruction error).

Nature makes a myriad form of each.