ComplexityNet: Increasing LLM Inference Efficiency by Learning Task Complexity

TL;DR

ComplexityNet is a fine-tuned model that accurately predicts task complexity, enabling more efficient LLM inference by reducing resource use while maintaining high accuracy.

Contribution

The paper introduces ComplexityNet, the first model to predict task complexity for LLMs, improving efficiency and balancing accuracy and resource consumption.

Findings

ComplexityNet achieved 79% accuracy in complexity prediction.

It reduced computational resources by 90%.

Maintained 86.7% code generation accuracy.

Abstract

We present ComplexityNet, a streamlined language model designed for assessing task complexity. This model predicts the likelihood of accurate output by various language models, each with different capabilities. Our initial application of ComplexityNet involves the Mostly Basic Python Problems (MBPP) dataset. We pioneered the creation of the first set of labels to define task complexity. ComplexityNet achieved a notable 79% accuracy in determining task complexity, a significant improvement over the 34% accuracy of the original, non fine-tuned model. Furthermore, ComplexityNet effectively reduces computational resource usage by 90% compared to using the highest complexity model, while maintaining a high code generation accuracy of 86.7%. This study demonstrates that fine-tuning smaller models to categorize tasks based on their complexity can lead to a more balanced trade-off between accuracy and efficiency in the use of Large Language Models. Our findings suggest a promising direction for optimizing LLM applications, especially in resource-constrained environments.