Paperopenalex
按 token 的是/否门控:从组合结构到真实减算
何淇 · Zenodo (CERN European Organization for Nuclear Research) 2026 · 2026
DOI: 10.5281/zenodo.23720312
0citations
Advanced Neural Network ApplicationsTopic ModelingNatural Language Processing TechniquesSecurity tokenComputer scienceToken passing+2
Zenodo (CERN European Organization for Nuclear Research)RepositoryCitations by source
Synced 2026-10-31T02:32:09.482604Z0OpenAlex · primary0Semantic Scholar0Crossref
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Abstract
【工作稿·请勿引用】本篇方向可能是对的,但多处实验与结论尚未复核,已暂时从系列合集中撤出,待完善后另行发布。内容:把"判定 J"实现为每层每 token 的 0/1 门;合成实验(层门/跨层门/按需迭代/空间-时间前沿);GPT-2 上的模拟门控与真实减算;预算版;文末附"待复核清单"(10 条)。全部为 P 侧工程实验,不声称生命属性。
Related papers
- 层的路径与冗余:GPT-2 上的层扰动实验
Zenodo (CERN European Organization for Nuclear Research) 2026 · 1 citations
- 语言结构作为自变量:GPT-2 上的三组实验
Zenodo (CERN European Organization for Nuclear Research) 2026 · 0 citations
Discussion (0)
No comments yet — start the discussion below.