The increasing demand for durable and efficient corrosion inhibitors has intensified interest in macrocyclic ligands as an emerging class of tunable corrosion inhibitors. Several salient features, including macrocyclic effect, host-guest recognition, preorganized multidentate coordination, extensive surface coverage, and metal-ion complexation within a single molecular framework, make them an attractive and effective class of inhibitors. These properties also make them different from traditional corrosion inhibitors. The present review provides broad coverage of phthalocyanines, crown ethers, porphyrins, azamacrocycles, and calixarenes, with particular focus on their structure-performance relationships, coordination mechanisms, adsorption, and the formation of corrosion-protective films on different substrates, including steel, iron, aluminum alloys, copper, and magnesium in acidic, alkaline, and saline environments. Macrocyclic inhibitors exhibit a broad range of inhibition efficiencies depending on their molecular structure, metallic substrate, and experimental conditions. This review also highlights the roles of donor atoms, molecular geometry, ring or cavity size, metal coordination, substituents, and solubility in the adsorption, coordination, and corrosion inhibition of macrocyclic ligands. This article also discusses their challenges related to poor solubility, synthetic complexity, scalability, environmental consideration, and limited industrial-scale validation under realistic conditions. Lastly, research gaps in macrocyclic-based corrosion inhibition have also been presented with outlines, enabling effective and sustainable design of macrocycle-based corrosion inhibitors.
Fuzzy rough set, as an important mathematical method, plays a vital role in feature selection of numerical data. Most existing fuzzy rough set-based feature selection methods merely consider a single-type feature or are mainly designed for datasets with complete decision labels. Additionally, since the acquisition of object labels requires expensive time and resource costs, only some objects possess decision labels. Partially labeled hybrid data consisting of symbolic, numerical and missing features are often more common in practical applications. In this study, we propose a feature selection method based on fuzzy information granularity for partially labeled hybrid data, which encompasses three key designs. Firstly, an improved fuzzy decision strategy based on the neighborhood granule is proposed to incorporate the potential decision label besides complete decision labels. Secondly, a monotonic fuzzy information granularity measure is designed to simultaneously assess the fuzziness and uncertainty of feature subsets in partially labeled hybrid data. Based on this, a heuristic feature selection algorithm is proposed to select the optimal feature subset. Thirdly, an accelerator mechanism is proposed to progressively shrink the object space to further reduce the running time of the proposed feature selection algorithm. The extensive experimental results demonstrate that the proposed method improves classification performance while reducing time cost.
The ability of large language models (LLMs) to follow instructions is crucial for their practical applications, yet the underlying mechanisms remain poorly understood. This paper presents a novel framework that leverages sparse autoencoders (SAE) to interpret how instruction following works in these models. We demonstrate how the features we identify can effectively steer model outputs to align with given instructions. Through analysis of SAE latent activations, we identify specific latents responsible for instruction following behavior. Our findings reveal that instruction following capabilities are encoded by a distinct set of instruction-relevant SAE latents. These latents both show semantic proximity to relevant instructions and demonstrate causal effects on model behavior. Our research highlights several crucial factors for achieving effective steering performance: precise feature identification, the role of final layer, and optimal instruction positioning. Additionally, we demonstrate that our methodology scales effectively across SAEs and LLMs of varying sizes.
Large vision-language models (LVLMs) have made significant progress in chart understanding. However, financial charts, characterized by complex temporal structures and domain-specific terminology, remain notably underexplored. We introduce FinChart-Bench, the first benchmark specifically focused on real-world financial charts. FinChart-Bench comprises 1,200 financial chart images collected from 2015 to 2024, each annotated with True/False (TF), Multiple Choice (MC), and Question Answering (QA) questions, totaling 7,016 questions. We conducted a comprehensive evaluation of 26 state-of-the-art LVLMs on FinChart-Bench. Our evaluation reveals critical insights: (1) the performance gap between open-source and closed-source models is narrowing, (2) performance degradation occurs in upgraded models within families, (3) many models struggle with instruction following, (4) both advanced models show significant limitations in spatial reasoning abilities, and (5) current LVLMs are not reliable enough to serve as automated evaluators. These findings highlight important limitations in current LVLM capabilities for financial chart understanding.
Large Language Models (LLMs) excel at many reasoning tasks but struggle with knowledge-intensive queries due to their inability to dynamically access up-to-date or domain-specific information. Retrieval-Augmented Generation (RAG) has emerged as a promising solution, enabling LLMs to ground their responses in external sources. However, existing RAG methods lack fine-grained control over both the query and source sides, often resulting in noisy retrieval and shallow reasoning. In this work, we introduce DeepSieve, an agentic RAG framework that incorporates information sieving via LLM-as-a-knowledge-router. DeepSieve decomposes complex queries into structured sub-questions and recursively routes each to the most suitable knowledge source, filtering irrelevant information through a multi-stage distillation process. Our design emphasizes modularity, transparency, and adaptability, leveraging recent advances in agentic system design. Experiments on multi-hop QA tasks across heterogeneous sources demonstrate improved reasoning depth, retrieval precision, and interpretability over conventional RAG approaches.
