在本文中,我们报告了证据表明,GPT 中的事实关联对应于可以直接编辑的本地化计算。 大型语言模型可以预测有关世界的事实陈述(Petroni 等人,2019 年;Jiang 等人,2020 年;Roberts 等人,2020 年)。例如,给定前缀“The Space Needle is located in the city of”,GPT 将可靠地预测真实答案:“Seattle”(图 1a)。
3 Interventions on Weights for Understanding Factual Association Storage 3.1 Rank-One Model Editing: Viewing the Transformer MLP as an Associative Memory 3.2 Evaluating ROME: Zero-Shot Relation Extraction (zsRE) 泻药。对于GPT族模型的internal interpretation感兴趣的话,可以从下面这篇ROME开始。这篇同样也是...
We analyze the storage and recall of factual associations in autoregressive transformer language models, finding evidence that these associations correspond to localized, directly-editable computations. We first develop a causal intervention for identifying neuron activations that are decisive in a model's ...
This repository provides an implementation of Rank-One Model Editing (ROME) on auto-regressive transformers (GPU-only). We currently support OpenAI's GPT-2 XL (1.5B) and EleutherAI's GPT-J (6B). The release of a 20B GPT-like model from EleutherAI is expected soon; we hope to support...
网站:Locating and Editing Factual Associations in GPT 一、简介二、干预激活,追踪信息流--- 2.1 事实关联的因果追踪--- 2.2 因果追踪结果--- 2.3 局部事实关联假说三、了解事实关联存储权重的干预措施--- 3.1 Rank-One 模型编辑:将transformer MLP 视为关联记忆体--- 3.2 评估 ROME:zero-shot关系抽取(zsRE...