Semantic Landauer Extension: Semantic Entropy Energy Cost Dataset
收藏资源简介：
This dataset supports the paper “BERT-Based Semantic Entropy under Landauer’s Principle: A Framework for Quantifying Text Processing Energy Cost.” It includes: (1) per-sentence BERT attention distributions (JSON); (2) computed semantic entropy values SsemS_{\mathrm{sem}}Ssem (CSV); (3) normalized energy values EnormE_{\mathrm{norm}}Enorm (CSV); (4) hardware energy estimation data for NVIDIA A100 GPU and Intel Loihi neuromorphic chip (CSV); (5) GPT-3 API energy cost and electricity pricing examples (CSV); (6) plotting scripts and Jupyter notebook to reproduce all figures; (7) Flask demo server and simple web UI code. Authors: PS BigBig (Independent Developer and Researcher) Contact: hello@onestardao.com. All papers: https://onestardao.com/papers
