Zero-shot object navigation tasks agents with locating target objects in unseen environments—a core capability of embodied intelligence. While recent vision-language navigation methods leverage Large Language Models (LLMs) for multimodal reasoning, they suffer from two key limitations: (1) semantic misalignment between language-grounded maps and real-world layouts, and (2) inefficiency due to LLMs’ lack of specialization for navigation-specific tasks. To address these challenges, we propose Chain-of-Search (CoS), a novel parameter-efficient framework that enables human-like decision-making via iterative semantic reasoning. First, CoS replaces traditional global maps with an optimal-benefit multi-map construction that continuously balances expected gain and cost throughout the navigation process. Second, we introduce a Parameter-Efficient Intent Aligner (PEIA), trained via a prompt-guided paradigm to align directional decisions with navigation intent. PEIA injects semantic cues into benefit-aware maps, enabling more rational and goal-consistent exploration. Finally, a Reflection-Guided Destination Verifier (RDV) confirms whether the target is reached via language-driven reasoning and corrects potential errors through self-reflection. CoS achieves state-of-the-art performance on HM3D (+2.8% SR) and MP3D (+1.2% SR) without relying on LLMs, demonstrating the effectiveness of lightweight, reasoning-centered navigation.

Infrared small target segmentation technology plays an important role in fields such as missile warning, maritime rescue, and military reconnaissance. However, CNN methods based on convolution tend to lose information regarding infrared small targets, resulting in poor segmentation performance. On the other hand, methods based on transformers, lacking convolution-induced biases, also struggle to achieve good results. To address this issue, this article proposes a model called Multiscale Feature Fusion Spatial-channel Attention Network (MFFSANet) for the segmentation of infrared small targets. The MFFSANet model consists of three blocks: the Multi-scale Convolution Fusion Attention (MCFA) block, the Hierarchical Guided Channel Attention (HGCA) block, and the Atrous Residual U-Block (ARU). The MCFA block leverages multi-scale atrous convolutions and self-attention mechanisms to obtain both local and global information about the image, learning the difference between target features and background noise features, thus enabling the model to suppress background noise in infrared images. The HGCA block leverages coarser information to guide the learning of finer features, assigning weights to decisive channels, and reducing redundant information. This reduces background noise in infrared images, making small targets stand out more clearly against the background. The ARU facilitates interaction between feature maps of different layers and scales, enabling the model to recognize the characteristics of small infrared targets in a more detailed and comprehensive manner. Extensive experiments conducted on four publicly available datasets, namely SIRST, IRSTD-1k, NUDT-SIRST, and SIRST-Aug, demonstrate the effectiveness and superiority of the proposed MFFSANet method compared to several SOTA infrared small target segmentation methods. The source code is available athttps://github.com/change68/MFFSANet.