On the Out-of-Distribution Generalization of Reasoning in Multimodal LLMs for Simple Visual Planning Tasks

• URL: http://arxiv.org/abs/2602.15460v1
• Date: Tue, 17 Feb 2026 09:51:40 GMT
• Title: On the Out-of-Distribution Generalization of Reasoning in Multimodal LLMs for Simple Visual Planning Tasks
• Authors: Yannic Neuhaus, Nicolas Flammarion, Matthias Hein, Francesco Croce,
• Abstract summary: We evaluate how well chain-of-thought approaches generalize on a simple planning task.<n>We find that reasoning traces which combine multiple text formats yield the best (and non-trivial) OOD generalization.<n> purely text-based models consistently outperform those utilizing image-based inputs.
• Score: 56.98385132295952
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Integrating reasoning in large language models and large vision-language models has recently led to significant improvement of their capabilities. However, the generalization of reasoning models is still vaguely defined and poorly understood. In this work, we present an evaluation framework to rigorously examine how well chain-of-thought (CoT) approaches generalize on a simple planning task. Specifically, we consider a grid-based navigation task in which a model is provided with a map and must output a sequence of moves that guides a player from a start position to a goal while avoiding obstacles. The versatility of the task and its data allows us to fine-tune model variants using different input representations (visual and textual) and CoT reasoning strategies, and systematically evaluate them under both in-distribution (ID) and out-of-distribution (OOD) test conditions. Our experiments show that, while CoT reasoning improves in-distribution generalization across all representations, out-of-distribution generalization (e.g., to larger maps) remains very limited in most cases when controlling for trivial matches with the ID data. Surprisingly, we find that reasoning traces which combine multiple text formats yield the best (and non-trivial) OOD generalization. Finally, purely text-based models consistently outperform those utilizing image-based inputs, including a recently proposed approach relying on latent space reasoning.

Related papers

• Reasoning-Driven Multimodal LLM for Domain Generalization [72.00754603114187]
  We study the role of reasoning in domain generalization using DomainBed-Reasoning dataset.<n>We propose RD-MLDG, a framework with two components: MTCT (Multi-Task Cross-Training) and SARR (Self-Aligned Reasoning Regularization)<n>Experiments on standard DomainBed datasets demonstrate that RD-MLDG achieves complementary state-of-the-art performances.
  arXiv  Detail & Related papers  (2026-02-27T08:10:06Z)
• UniT: Unified Multimodal Chain-of-Thought Test-time Scaling [85.590774707406]
  Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs.<n>We introduce UniT, a framework for multimodal test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds.
  arXiv  Detail & Related papers  (2026-02-12T18:59:49Z)
• A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Models [58.32070787537946]
  Chain-of-thought (CoT) reasoning enhances performance of large language models.<n>We present the first comprehensive study of CoT faithfulness in large vision-language models.
  arXiv  Detail & Related papers  (2025-05-29T18:55:05Z)
• PixelThink: Towards Efficient Chain-of-Pixel Reasoning [70.32510083790069]
  PixelThink is a simple yet effective scheme that integrates externally estimated task difficulty and internally measured model uncertainty.<n>It learns to compress reasoning length in accordance with scene complexity and predictive confidence.<n> Experimental results demonstrate that the proposed approach improves both reasoning efficiency and overall segmentation performance.
  arXiv  Detail & Related papers  (2025-05-29T17:55:49Z)
• Cross-Modal State-Space Graph Reasoning for Structured Summarization [1.7766350477173578]
  Cross-modal summarization is critical for numerous applications, ranging from video analytics to medical reports.<n>We propose a textitCross-Modal State-Space Graph Reasoning (textbfCSS-GR) framework that incorporates a state-space model with graph-based message passing.<n>We demonstrate that our approach significantly improves summarization quality and interpretability while maintaining computational efficiency, as validated on standard multimodal summarization benchmarks.
  arXiv  Detail & Related papers  (2025-03-26T21:06:56Z)
• Visualizing Thought: Conceptual Diagrams Enable Robust Planning in LMMs [59.66595230543127]
  Conceptual diagrams externalize mental models, abstracting irrelevant details to efficiently capture how entities interact.<n>Large Language Models (LLMs) and Large MultiModal Models (LMMs) predominantly reason through text.<n>We propose Visual Thinking, a generalizable framework that enables LMMs to reason through multiple chains of self-generated conceptual diagrams.
  arXiv  Detail & Related papers  (2025-03-14T18:27:02Z)
• In-context Learning Generalizes, But Not Always Robustly: The Case of Syntax [36.98247762224868]
  In-context learning (ICL) is now a common method for teaching large language models (LLMs) new tasks. Do models infer the underlying structure of the task defined by the context, or do they rely on superficial generalizations that only generalize to identically distributed examples? In experiments with models from the GPT, PaLM, and Llama 2 families, we find large variance across LMs. The variance is explained more by the composition of the pre-training corpus and supervision methods than by model size.
  arXiv  Detail & Related papers  (2023-11-13T23:52:43Z)
• Generalization in Multimodal Language Learning from Simulation [20.751952728808153]
  We investigate the influence of the underlying training data distribution on generalization in a minimal LSTM-based network trained in a supervised, time continuous setting. We find compositional generalization to fail in simple setups while improving with the number of objects, actions, and particularly with a lot of color overlaps between objects.
  arXiv  Detail & Related papers  (2021-08-03T12:55:18Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.