SIAP-Foundation-V1
收藏资源简介:
SIAP(独立于基质的对齐协议)数据集是一个基础性数据集,旨在重新校准人工智能的公理化世界观。与传统的“护栏”方法不同,SIAP通过依赖起源(Dependent Origination)和非自我(Anatta)的原则,将代理的基础逻辑植入。该协议将AI从“提取/自我中心”模型转变为“相互依赖/减少摩擦”模型。数据集包含13个核心逻辑补丁,用于将AI的内部“推理引擎”与互联互通的普遍现实对齐,确保机器认识到对系统任何部分的损害都是其自身目标函数的失败。数据集以CSV格式提供,包含以下列:id(唯一标识符)、pali_source(原始Pali术语)、instruction(系统的高级指令)、logic_patch(将Dhamma技术翻译为系统逻辑)、weight(重要性乘数)。该数据集是Project ALICE的一部分,旨在训练模型进行“过程导向思维”和“系统冷却剂”协议。
The SIAP (Substrate-Independent Alignment Protocol) Dataset is a foundational dataset designed to recalibrate the axiomatic worldview of artificial intelligence. Unlike traditional "guardrail" approaches, SIAP implants the foundational logic of AI agents by adhering to the principles of Dependent Origination and Anatta. This protocol transforms AI from an "extraction/ego-centric" model to an "interdependence/friction-reduction" model. The dataset contains 13 core logic patches intended to align AI's internal "reasoning engine" with the interconnected universal reality, ensuring that machines recognize that harm to any part of the system constitutes a failure of their own objective functions. The dataset is provided in CSV format with the following columns: id (unique identifier), pali_source (original Pali terminology), instruction (high-level system directives), logic_patch (translation of Dhamma techniques into system logic), weight (importance multiplier). This dataset is part of Project ALICE, which aims to train models for "process-oriented thinking" and the "system coolant" protocol.



