Papers by Taemin Ha
Large Language Models as Realistic Microservice Trace Generators (2025.emnlp-main)
Copied to clipboard
| Challenge: | Obtaining real-world traces is difficult due to limited public data availability and the difficulty of collecting them at large scale from diverse environments. |
| Approach: | They propose to train a large language model to generate microservice call graphs using a recursive approach to capture hierarchical structures and implicit constraints in such traces. |
| Outcome: | The proposed method outperforms existing methods in accuracy and validity. |