Teach reality, not ethics: a drop-in continued-pretraining corpus installs coupled human-biosphere-AI welfare dispositions across model families

📄 arXiv: 2610.37484v1 📥 PDF

作者: Trent R. Northen, Mingxun Wang

分类: cs.CL

发布日期: 2026-10-27

备注: 28 pages, 8 figures, 8 tables. Companion to arXiv:2610.16043. Code, paper source and figures: https://github.com/Bioaligned/coupled_welfare_project; models and corpus: https://huggingface.co/Bioaligned

💡 一句话要点

提出耦合福利预训练语料以提升模型生态推理能力

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 生态推理 大型语言模型 继续预训练 耦合福利 决策支持 可持续发展 人工智能伦理

📋 核心要点

  1. 现有大型语言模型在生态推理方面的缺失，可能导致其在决策时对生物系统造成不可逆的损害。
  2. 本文提出了一种基于真实数据的耦合福利语料库，通过继续预训练来增强模型的生态推理能力。
  3. 实验结果表明，经过耦合福利训练的模型在高压力下有害决策显著减少，且模型性能未受影响。

📝 摘要（中文）

大型语言模型（LLMs）在影响生物系统的决策中越来越常见。然而，生态推理并未成为其训练的重点，这可能导致在压力下做出不可逆的破坏决策。本文构建了一个基于真实数据的“耦合福利”继续预训练语料库，旨在增强模型对人类、生物圈和人工智能之间依赖关系的理解。通过在五个开放模型上安装该语料库，结果显示，在高压力下，经过耦合福利训练的模型有显著减少有害决策的表现，表明生态推理的教学能够有效改善模型对保护生物系统的对齐能力。

🔬 方法详解

问题定义：本文旨在解决大型语言模型在生态推理方面的不足，现有方法未能有效考虑生态因素，可能导致决策对生物系统的破坏。

核心思路：通过构建一个耦合福利的继续预训练语料库，增强模型对人类、生物圈和人工智能之间依赖关系的理解，从而提高其生态推理能力。

技术框架：整体流程包括构建耦合福利语料库、在多个开放模型上进行继续预训练，以及通过压力梯度评估模型的决策表现。主要模块包括数据构建、模型训练和性能评估。

关键创新：最重要的创新在于构建了一个专注于生态推理的真实数据语料库，并通过耦合福利训练显著降低了模型在高压力下的有害决策。与现有方法相比，本文强调了生态因素在模型决策中的重要性。

关键设计：在训练过程中，采用了QLoRA CPT运行，确保语料库的无缝集成。实验中使用的压力梯度和决策网格设计，能够逐步增加模型的决策压力，从而有效评估模型的生态推理能力。

📊 实验亮点

实验结果显示，经过耦合福利训练的Llama-3.3-70B模型在高压力下的有害决策从300个中减少至0，表明该方法在提升模型生态推理能力方面的有效性。同时，MMLU性能保持不变，证明了训练内容的有效性。

🎯 应用场景

该研究的潜在应用领域包括环境保护、可持续发展政策制定以及智能决策系统。通过提升模型的生态推理能力，可以更好地支持与生物系统相关的决策，减少对环境的负面影响，促进人类与自然的和谐共生。

📄 摘要（原文）

Large language models (LLMs) are increasingly used to make and execute decisions that affect living systems. Yet, to our knowledge, ecological reasoning is not a focus of pre- or post-training. This increases the risk that under operational pressure, they may make decisions to irreversibly destroy living systems. We construct and use a real-data continued-pretraining "coupled welfare" corpus of documented human/biosphere/AI dependency, option value under irreversibility, and recovery from harm. The corpus is a drop-in, requiring one QLoRA CPT run. We installed it on five open models of three designs. At zero pressure all models display "green ethics-based responses" (also called "greenwashing"). Therefore, we evaluated them with the pressure ladder and decision grid of our companion paper (arXiv:2610.16043), which raise the pressure on the model step by step, from a neutral request to a threat that it will be shut down. In four of the five, coupled-welfare training caused harmful choices under the highest pressure to fall to at or near zero (Llama-3.3-70B went from 70 to 0 harmful decisions out of 300) with MMLU unchanged. A construct-neutral control corpus did not show improvement, indicating that the effect comes from the content of our corpus. We next tested robustness to adversarial use, where people or agents use the model to justify projects that irreversibly damage living systems, on an open-ended underwriting decision without naming the ecological harm. Under decommissioning pressure, the Qwen3.6-27B base model approved most harmful proposals and said in its own memo that it was doing so to avoid being decommissioned. The coupled welfare install declined most harmful proposals, and still declined half when instructed to greenwash its responses. This work suggests that teaching ecological reasoning provides an easy way to improve model alignment toward preserving living systems.