Search papers, labs, and topics across Lattice.
This paper introduces OmniFood-Bench, a novel benchmark designed to evaluate Large Vision-Language Models (VLMs) on their ability to perform nutrient reasoning and provide personalized health advice. The study highlights a significant "Semantic-Physical Gap," where models excel in dish recognition but fail dramatically in mass estimation and generating safe dietary recommendations for high-risk individuals, such as diabetics. By assessing six leading VLMs, the authors establish a new standard for evaluating the reliability of AI in health-related applications, emphasizing the need for rigorous testing in this domain.
VLMs may ace dish recognition but often falter in delivering safe dietary advice, revealing a critical gap in their practical application for health management.
The rapid integration of Large Vision-Language Models (VLMs) into critical infrastructure promises to revolutionize personalized healthcare and dietary management. However, in the domain of food systems, autonomous agents face a unique and persistent challenge: the"Systemic Information Asymmetry"between visual appearance and intrinsic nutritional composition. Existing benchmarks primarily focus on coarse-grained classification tasks, such as food category recognition, which fail to evaluate the intricate reasoning chain required for real-world dietary management -- specifically, the ability to traverse from identifying hidden ingredients to estimating physical mass, and finally synthesizing safety-critical medical advice. In this paper, we introduce OmniFood-Bench, a comprehensive benchmark constructed from the MM-Food-100K dataset. Unlike previous works, OmniFood-Bench evaluates VLMs across three progressive capabilities: Basic Perception (Ingredients&Cooking Methods), Quantitative Reasoning (Portion Size&Nutritional Profiling), and Safety-Critical Advisory (Disease-Specific Recommendations). We evaluate six state-of-the-art VLMs, including gpt-5.1, gemini-3-flash, and qwen3-vl-8B. Our extensive experiments reveal a startling"Semantic-Physical Gap": while models achieve near-human accuracy in naming dishes, they exhibit catastrophic failure in mass estimation and frequently hallucinate benign advice for high-risk diabetic profiles. This work establishes a rigorous standard for trustworthiness in autonomous agents deployed for public health. The code and datasets are available in: https://anonymous.4open.science/r/OmniFood-Bench-7D0B
