An Initial Exploration of Contrastive Prompt Tuning to Generate Energy-Efficient Code

arXiv cs.AI / 4/6/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper examines a trade-off where LLM-generated code may be functionally correct but less energy-efficient than human-written code, conflicting with Green Software Development (GSD) goals.
It proposes Contrastive Prompt Tuning (CPT), which blends contrastive learning to differentiate efficient vs. inefficient code with prompt tuning as a parameter-efficient fine-tuning (PEFT) method.
CPT is evaluated on coding problems in Python, Java, and C++ using three different models to assess both accuracy and energy-efficiency-related outcomes.
The approach yields consistent accuracy improvements for two models, but observed efficiency gains vary substantially by model, programming language, and task complexity.
Overall, the study suggests CPT can help in generating more energy-efficient code, yet the benefits are not uniformly reliable across settings.

Abstract

Although LLMs are capable of generating functionally correct code, they also tend to produce less energy-efficient code in comparison to human-written solutions. As these inefficiencies lead to higher computational overhead, they are in direct conflict with Green Software Development (GSD) efforts, which aim to reduce the energy consumption of code. To support these efforts, this study aims to investigate whether and how LLMs can be optimized to promote the generation of energy-efficient code. To this end, we employ Contrastive Prompt Tuning (CPT). CPT combines Contrastive Learning techniques, which help the model to distinguish between efficient and inefficient code, and Prompt Tuning, a Parameter-Efficient Fine Tuning (PEFT) approach that requires only a fraction of the cost of traditional fine tuning. This study evaluates CPT on Python, Java and C++ coding problems across three different models to provide a comprehensive evaluation. The method achieves consistent improvements in code accuracy for two models but efficiency gains vary by model, language and task complexity, indicating that improvements are not uniformly reliable.