analiza aiqbench
HA-VLN 2.0 introduces a unified benchmark and dataset that explicitly evaluates navigation agents on social-awareness constraints, including personal-space adherence and multi-human interactions. The release of simulators and an open leaderboard highlights a critical gap in current models, which suffer significant performance drops when faced with dynamic, crowded environments. This work provides a necessary foundation for advancing safe, human-centric robotics by validating sim-to-real transfer in socially grounded scenarios.
HA-VLN 2.0: An Open Benchmark and Leaderboard for Human-Aware Navigation in Discrete and Continuous Environments with Dynamic Multi-Human Interactions
arXiv:2503.14229v4 Announce Type: replace-cross Abstract: Vision-and-Language Navigation (VLN) has been studied mainly in either discrete or continuous spaces, with little attention to dynamic, crowded environments. We present HA-VLN 2.0, a unified benchmark introducing explicit social-awareness constraints. Our contributions are: (i) a standardized task and metrics capturing both goal accuracy and personal-space adherence; (ii) HAPS 2.0 dataset and simulators modeling multi-human interactions, outdoor contexts, and finer language-motion alignment; (iii) benchmarks on 16,844 socially grounded instructions, revealing sharp performance drops of leading agents under human dynamics and partial observability; and (iv) real-world robot experiments validating sim-to-real transfer, with an open leaderboard enabling transparent comparison. Results show that explicit social modeling improves navigation robustness and reduces collisions, underscoring necessity of human-centric approaches. By releasing datasets, simulators, baselines, and protocols, HA-VLN 2.0 provides a strong foundation for safe, human-aware navigation research.
Abstract:Vision-and-Language Navigation (VLN) has been studied mainly in either discrete or continuous spaces, with little attention to dynamic, crowded environments. We present HA-VLN 2.0, a unified benchmark introducing explicit social-awareness constraints. Our contributions are: (i) a standardized task and metrics capturing both goal accuracy and personal-space adherence; (ii) HAPS 2.0 dataset and simulators modeling multi-human interactions, outdoor contexts, and finer language-motion alignment; (iii) benchmarks on 16,844 socially grounded instructions, revealing sharp performance drops of leading agents under human dynamics and partial observability; and (iv) real-world robot experiments validating sim-to-real transfer, with an open leaderboard enabling transparent comparison. Results show that explicit social modeling improves navigation robustness and reduces collisions, underscoring necessity of human-centric approaches. By releasing datasets, simulators, baselines, and protocols, HA-VLN 2.0 provides a strong foundation for safe, human-aware navigation research.
Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO)
arXiv:2503.14229 [cs.AI]
(or arXiv:2503.14229v4 [cs.AI] for this version)
Text integral sub licența deschisă a publicației · 151 cuvinte
Subiecte conexe
HA-VLN 2.0: An Open Benchmark and Leaderboard for Human-Aware Navigation in Discrete and Continuous Environments with Dynamic Multi-Human Interactions
arXiv · 19 aug. 2026
State of the Field: AI for Epistemics and Coordination
LessWrong · 2 oct. 2026
News Archive 2026
Quantum Computing Report · 25 iul. 2026
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · 17 sept. 2026