Given the escalating challenge of global plastic pollution, pyrolysis has emerged as an important technology for converting waste plastics into high-value chemicals and liquid fuels. However, due to complex reaction networks, variable feedstock compositions, and multiphase flow characteristics, traditional methods face significant bottlenecks in process optimization and mechanistic exploration. Artificial intelligence (AI) offers a novel perspective for overcoming these complexities with its inherent advantages in processing high-dimensional data involving nonlinear interactions among variables. Against this backdrop, this review comprehensively surveys and analyzes the latest progress and limitations of applying AI to plastic pyrolysis, focusing on the identification of pyrolysis kinetic parameters using intelligent optimization algorithms and AI methods, the prediction of plastic pyrolysis reaction outcomes using AI, the exploration of intrinsic relationships between key process features and product distribution, and pyrolysis mechanisms using interpretability tools, and reactor optimization combining physical models such as computational fluid dynamics (CFD) with AI-based surrogate models. Furthermore, emerging applications of natural language processing (NLP) and large language models (LLMs) in automating the construction of standardized pyrolysis databases and knowledge graphs are explored. While summarizing technical advantages, current weaknesses such as data scarcity, insufficient model generalization, and the lack of physical information coupling are also analyzed and discussed. Finally, a conceptual architecture for an LLM-driven pyrolysis agent is proposed to integrate mechanism retrieval, process simulation, and experimental support. This review provides theoretical insights and future perspectives for intelligent process optimization and potential engineering applications of plastic pyrolysis for the conversion of waste plastics into fuels and chemicals.

Quantitative structure–property relationship (QSPR) modeling is a fundamental approach for property-oriented fuel design because it establishes quantitative relationships between molecular structures and physicochemical properties. Conventional QSPR models generally rely on manually engineered molecular descriptors, are usually developed to predict a single property, and often cannot readily incorporate newly available data, limiting their scalability and long-term applicability. Recent advances in large language models (LLMs) have demonstrated remarkable capabilities in text understanding and parsing knowledge in specialized fields such as chemistry. This study proposes FuelProp-LM, a QSPR framework based on instruction tuning and in-context learning that predicts multiple fuel properties with only the SMILES notation of a fuel molecule as input. Four open-source language models were fine-tuned using 100000 data entries obtained from the PubChem database. To enhance contextual reasoning, high-information prompts were constructed through MACCS molecular fingerprint similarity retrieval. The proposed framework was evaluated on 10 fuel physicochemical property datasets and compared with the widely-adopted LLM DeepSeek-V3.2 and existing QSPR approaches. FuelProp-LM consistently achieved competitive or superior predictive performance across downstream tasks. Quantitative analysis further revealed how instruction tuning and in-context learning improve molecular property prediction by enhancing chemically relevant attention patterns. The accuracy, generalizability, data scalability, and ease of use of this work demonstrate that FuelProp-LM can serve as a practical and flexible tool for fuel property prediction and play a significant role in future AI-assisted fuel design.

Researchers are deploying a suite of AI agents not to replace humans, but to relieve administrative burdens and ease patient suffering.