Carbon-Taxed Transformers: A Green Compression Pipeline for Overgrown Language Models

TL;DR

This paper introduces Carbon-Taxed Transformers (CTT), a systematic compression pipeline inspired by economic carbon taxation, significantly reducing the environmental impact and computational costs of large language models in software engineering tasks.

Contribution

The paper proposes a novel, principled compression pipeline for LLMs that balances efficiency and accuracy, inspired by economic carbon taxation principles.

Findings

Up to 49x memory reduction in models.

Inference time reduced by up to 10x in clone detection.

Up to 81% reduction in CO2 emissions.

Abstract

The accelerating adoption of Large Language Models (LLMs) in software engineering (SE) has brought with it a silent crisis: unsustainable computational cost. While these models demonstrate remarkable capabilities in different SE tasks, they are unmanageably large, slow to deploy, memory-intensive, and carbon-heavy. This reality threatens not only the scalability and accessibility of AI-powered SE, but also its long-term environmental sustainability. The research challenge is clear: we must go beyond accuracy and address efficiency and environmental cost as first-class design constraints. To meet this challenge, we introduce Carbon-Taxed Transformers (CTT), a systematic multi-architectural compression principled pipeline ordering inspired by economic carbon taxation principles. Drawing from the economic concept of carbon pricing, CTT operationalizes a computational carbon tax that penalizes architectural inefficiencies and rewards deployment-ready compression. We evaluate CTT across three core SE tasks: code clone detection, code summarization, and code generation, with models spanning encoder-only, encoder-decoder, and decoder-only architecture. Our results show that CTT delivers on inference: (1) up to 49x memory reduction, (2) time reduction up to 8-10x for clone detection, up to 3x for summarization, and 4-7x for generation, (3) up to 81% reduction in CO2 emissions and (4) CTT retains around 98% accuracy on clone detection, around 89% on summarization, and up to 91% (textual metrics) and 68% (pass@1) for generation. Two ablation studies show that pipeline ordering and individual component contributions are both essential, providing empirical justification for CTT's design and effectiveness. This work establishes a viable path toward responsible AI in SE through aggressive yet performance-preserving compression.