Reproduce “Tug of Watts: A study on the Energy - Accuracy Trade-off of Large Language Models on Medical Question Answering Tasks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Viju Sudhi, Giulio Billi, Andreea Scrob, Regina Stodden, Philipp Cimiano
The authors compare 13 language models on medical multiple-choice benchmarks using accuracy and inference energy. They evaluate pre-trained and LoRA fine-tuned Qwen, Llama, and Gemma models, using Pareto analysis and CIRC, a five-level joint rating. Fine-tuned Qwen3-14B achieves 81.0% average accuracy, with 187% more inference energy than Qwen3-0.6B; findings are benchmark-specific, not clinical validation.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.