Research map / Large language models / AI and society
Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning
Peter Henderson, Jieru Hu, Joshua Romoff, Emma Brunskill, Dan Jurafsky, Joelle Pineau
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use FPOs as a proxy for energy, across architectures | pretrained image classifiers · image classification inference | correlation with energy: R2 = 0.083, Pearson 0.289 | No clear effect |
| use FPOs as a proxy for runtime, across architectures | pretrained image classifiers · image classification inference | correlation with time: R2 = 0.005, Pearson -0.074 | No clear effect |
| use FPOs as a proxy for energy within the VGG family | VGG · image classification inference | correlation with energy: R2 = .999, Pearson 1.0 | Improved |
| use parameter count as a proxy for energy | pretrained image classifiers · image classification inference | correlation with energy: R2 = 0.002, Pearson -0.048 | No clear effect |
| estimate energy and carbon from partial information instead of full tracking | Pong PPO experiment · PongNoFrameskip-v4 | energy and carbon estimate accuracy | Hurt |
| run jobs in Quebec instead of Estonia | cloud training · carbon emissions | CO2eq emitted: 30x lower | Improved |
| use PPO instead of DQN | RL baselines · Pong, 5M timesteps | energy (kWh): 888 kWh saved | Improved |
| run translation model inference in Estonia rather than Quebec | two MT models · emission gap between models | g CO2eq: .09 -> 3.04 | Hurt |
Closest papers by meaning
- Carbon Footprint Reduction for Sustainable Data Centers in Real-Time (AAAI 2024 · similarity 0.88)
- Holistically Evaluating the Environmental Impact of Creating Language Models (ICLR 2025 · similarity 0.85)
- The ML.ENERGY Benchmark: Toward Automated Inference Energy Measurement and Optimization (NeurIPS 2025 · similarity 0.85)
- LLMCarbon: Modeling the End-to-End Carbon Footprint of Large Language Models (ICLR 2024 · similarity 0.84)
- The Hidden Joules: Evaluating the Energy Consumption of Vision Backbones for Progress Towards More Efficient Model Inference (ICML 2025 · similarity 0.84)
- Promoting Sustainable Web Agents: Benchmarking and Estimating Energy Consumption Through Empirical and Theoretical Analysis (AAAI 2026 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
