Training large language models faces frequent interruptions due to various faults, demanding robust fault-tolerance. Existing backup-free methods, such as redundant computation, dynamic parallelism, and data rerouting, each incur performance penalties, whether from ongoing overhead, lengthy reconfigurations, or post-recovery inefficiencies. We propose Chameleon, an adaptive fault-tolerant system that intelligently selects optimal recovery strategies when a failure occurs. Chameleon achieves this through a unified performance model, expedient execution plan search, accurate performance estimation, and efficient communication optimizations. Experiments on a 32-card cluster show that Chameleon maintains a performance gap of within 11.00% between post-recovery and failure-free training, while preserving model convergence and efficient memory usage. Compared to state-of-the-art methods, Chameleon achieves up to 1.229x and 1.355x higher average throughput than Oobleck and Recycle, respectively.

Standard recommender systems usually rely only on past user ratings as well as optional profiles of customers and products. In e-commerce settings, however, a more complete understanding of the corresponding bi-directional impact between the demands of customers and the supply capabilities of providers can be the key to success. This motivates us to design a recommendation model that explicitly reflects the supply and demand chains. We propose a Multi-relational Coupled Tensor and Matrix Factorization model, which jointly models user ratings as well as supply chain relationships for product recommendation. In addition, our model can predict the links between suppliers and manufacturers. We design an algorithm based on the Alternating Direction Method of Multipliers (ADMM) technique. Experiments on real-world datasets find that the proposed model outperforms traditional methods.