Meanings and Measurements: Multi-Agent Probabilistic Grounding for Vision-Language Navigation — ThinkLLM