Beyond Interpretability: Exploring the Comprehensibility of Adaptive Video Streaming through Large Language Models
This addresses the problem of making adaptive video streaming algorithms more comprehensible for developers, representing an incremental advancement in human-AI collaboration for algorithm design.
The paper tackles the challenge that interpretability in adaptive video streaming algorithms doesn't guarantee developers' subjective comprehensibility, and introduces ComTree, a framework that generates decision trees meeting performance requirements and uses large language models to evaluate them for comprehensibility, ultimately selecting solutions that facilitate human understanding while maintaining competitive performance.
Over the past decade, adaptive video streaming technology has witnessed significant advancements, particularly driven by the rapid evolution of deep learning techniques. However, the black-box nature of deep learning algorithms presents challenges for developers in understanding decision-making processes and optimizing for specific application scenarios. Although existing research has enhanced algorithm interpretability through decision tree conversion, interpretability does not directly equate to developers' subjective comprehensibility. To address this challenge, we introduce \texttt{ComTree}, the first bitrate adaptation algorithm generation framework that considers comprehensibility. The framework initially generates the complete set of decision trees that meet performance requirements, then leverages large language models to evaluate these trees for developer comprehensibility, ultimately selecting solutions that best facilitate human understanding and enhancement. Experimental results demonstrate that \texttt{ComTree} significantly improves comprehensibility while maintaining competitive performance, showing potential for further advancement. The source code is available at https://github.com/thu-media/ComTree.