# Before We Trust Them: Decision-Making Failures in Navigation of Foundation Models

arXiv: https://arxiv.org/abs/2601.05529 · HTML: https://mlsanity.com/paper/2601.05529/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 35 references that are in the map.
- First posted on arXiv 9 Jan 2026.

## Builds on (most cited first, 23 of 23)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models](https://mlsanity.com/paper/2504.10479/) (2025)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting](https://mlsanity.com/paper/2305.04388/) (2023)
- [Towards Understanding Sycophancy in Language Models](https://mlsanity.com/paper/2310.13548/) (2023)
- [SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities](https://mlsanity.com/paper/2401.12168/) (2024)
- [Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces](https://mlsanity.com/paper/2412.14171/) (2024)
- [Beyond the Nav-Graph: Vision-and-Language Navigation in Continuous Environments](https://mlsanity.com/paper/2004.02857/) (2020)
- [Room-Across-Room: Multilingual Vision-and-Language Navigation with Dense Spatiotemporal Grounding](https://mlsanity.com/paper/2010.07954/) (2020)
- [SafeAgentBench: A Benchmark for Safe Task Planning of Embodied LLM Agents](https://mlsanity.com/paper/2412.13178/) (2024)
- [General Evaluation for Instruction Conditioned Navigation using Dynamic Time Warping](https://mlsanity.com/paper/1907.05446/) (2019)
- [Sycophancy in Large Language Models: Causes and Mitigations](https://mlsanity.com/paper/2411.15287/) (2024)
- [OpenFMNav: Towards Open-Set Zero-Shot Object Navigation via Vision-Language Foundation Models](https://mlsanity.com/paper/2402.10670/) (2024)
- [LangNav: Language as a Perceptual Representation for Navigation](https://mlsanity.com/paper/2310.07889/) (2023)
- [IS-Bench: Evaluating Interactive Safety of VLM-Driven Embodied Agents in Daily Household Tasks](https://mlsanity.com/paper/2506.16402/) (2025)
- [Self-Prompting Large Language Models for Zero-Shot Open-Domain QA](https://mlsanity.com/paper/2212.08635/) (2022)
- [A Framework for Benchmarking and Aligning Task-Planning Safety in LLM-Based Embodied Agents](https://mlsanity.com/paper/2504.14650/) (2025)
- [SafeMERGE: Preserving Safety Alignment in Fine-Tuned Large Language Models via Selective Layer-Wise Model Merging](https://mlsanity.com/paper/2503.17239/) (2025)
- [Evaluating Large Language Models on Spatial Tasks: A Multi-Task Benchmarking Study](https://mlsanity.com/paper/2408.14438/) (2024)
- [Decipher the Modality Gap in Multimodal Contrastive Learning: From Convergent Representations to Pairwise Alignment](https://mlsanity.com/paper/2510.03268/) (2025)
- [FoREST: Frame of Reference Evaluation in Spatial Reasoning Tasks](https://mlsanity.com/paper/2502.17775/) (2025)

## Abstract (arXiv)

High success rates on navigation-related tasks do not necessarily translate into reliable decision making by foundation models. To examine this gap, we evaluate current models on six diagnostic tasks spanning three settings: reasoning under complete spatial information, reasoning under incomplete spatial information, and reasoning under safety-relevant information. Our results show that the current metrics may not capture critical limitations of the models and indicate good performance, underscoring the need for failure-focused analysis to understand model limitations and guide future progress. In a path-planning setting with unknown cells, GPT-5 achieved a high success rate of 93%; Yet, the failed cases exhibit fundamental limitations of the models, e.g., the lack of structural spatial understanding essential for navigation. We also find that newer models are not always more reliable than their predecessors on this end. In reasoning under safety-relevant information, Gemini-2.5 Flash achieved only 67% on the challenging emergency-evacuation task, underperforming Gemini-2.0 Flash, which reached 100% under the same condition. Across all evaluations, models exhibited structural collapse, hallucinated reasoning, constraint violations, and unsafe decisions. These findings show that foundation models still exhibit substantial failures in navigation-related decision making and require fine-grained evaluation before they can be trusted.