analiza aiqbench
This research addresses critical safety challenges in autonomous navigation by quantifying semantic ambiguity within Vision-Language Models using Gaussian Context Distributions. By grounding open-vocabulary predictions on a continuous physical scale, the method generates dense uncertainty maps that correlate with real-world hazards like mixed terrain and visual artifacts. The approach offers a statistically rigorous framework for improving the reliability of VLM-driven systems in complex outdoor environments.
Estimating Semantic Ambiguity via Gaussian Context Distributions for VLM-Driven Traversability Analysis
arXiv:2609.08583v1 Announce Type: new Abstract: Autonomous navigation in unstructured environments requires robust scene understanding, yet Vision-Language Models (VLMs) often suffer from semantic ambiguity, where conflicting predictions can lead to dangerous failures. To address this, we present a novel pipeline for vision-based traversability estimation that explicitly models contextual uncertainty. Our approach utilizes Conceptual Anchoring to ground open-vocabulary VLM predictions onto a continuous physical traversability scale. By formulating the model's responses as a Gaussian Context Distribution (GCD), we derive both a dense traversability map and a dense uncertainty map based on the statistical properties of the distribution. Experimental validation on the real-world GOOSE dataset demonstrates that our proposed uncertainty metric effectively correlates with sources of ambiguity, such as visual artifacts and mixed terrain overlap. The method exhibits competitive performance while offering the distinct advantage of providing statistical uncertainty estimates to address semantic ambiguity, enabling safer and more reliable autonomous behavior in complex outdoor settings.
Abstract:Autonomous navigation in unstructured environments requires robust scene understanding, yet Vision-Language Models (VLMs) often suffer from semantic ambiguity, where conflicting predictions can lead to dangerous failures. To address this, we present a novel pipeline for vision-based traversability estimation that explicitly models contextual uncertainty. Our approach utilizes Conceptual Anchoring to ground open-vocabulary VLM predictions onto a continuous physical traversability scale. By formulating the model's responses as a Gaussian Context Distribution (GCD), we derive both a dense traversability map and a dense uncertainty map based on the statistical properties of the distribution. Experimental validation on the real-world GOOSE dataset demonstrates that our proposed uncertainty metric effectively correlates with sources of ambiguity, such as visual artifacts and mixed terrain overlap. The method exhibits competitive performance while offering the distinct advantage of providing statistical uncertainty estimates to address semantic ambiguity, enabling safer and more reliable autonomous behavior in complex outdoor settings.
Robotics (cs.RO)
arXiv:2609.08583 [cs.RO]
(or arXiv:2609.08583v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.08583 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 240 cuvinte
Subiecte conexe
Act, Think or Abstain: Complexity-Aware Adaptive Inference for Vision-Language-Action Models
arXiv · 28 iul. 2026
Robust Chance-Constrained Optimization using a Continuous Parameter Space Wasserstein-2 Ambiguity Set of Gaussian Mixtures
arXiv · 21 iul. 2026
Approximate Relative Entropy Constraints for Nonlinear Covariance Steering Under Distribution Ambiguity
arXiv · 21 iul. 2026
Breaking Time: A Fully Gaussian Framework for Distributed and Continuous-Time SLAM
arXiv · 5 iun. 2026