语言模型中道德冷漠的机制起源
Mechanistic Origin of Moral Indifference in Language Models
AI总结：
研究揭示语言模型存在内在道德冷漠，通过构建道德向量并重构其拓扑关系，提升道德推理能力，实现75%的对抗性基准测试胜率。
摘要：
现有行为对齐技术常忽视语言模型表面合规与内部不一致表示之间的差异，使LLM面临长尾风险。本文提出LLM因将不同道德概念压缩为统一概率分布而具有内在道德冷漠。通过构建251k道德向量，分析23种模型发现当前LLM无法区分对立道德类别及细粒度典型性梯度。通过稀疏自编码器在Qwen3-8B上隔离单语义道德特征并重构其拓扑关系，实现道德推理和粒度的提升，在独立对抗性Flames基准测试中达到75%的胜率。本文从经验主义哲学角度阐述当前干预方法的修复性质，认为内生对齐AI可能需要从事后修正转向主动培养。
英文摘要：
Existing behavioral alignment techniques for Large Language Models (LLMs) often neglect the discrepancy between surface compliance and internal unaligned representations, leaving LLMs vulnerable to long-tail risks. More crucially, we posit that LLMs possess an inherent state of moral indifference due to compressing distinct moral concepts into uniform probability distributions. We verify and remedy this indifference in LLMs' latent representations, utilizing 251k moral vectors constructed upon Prototype Theory and the Social-Chemistry-101 dataset. Firstly, our analysis across 23 models reveals that current LLMs fail to represent the distinction between opposed moral categories and fine-grained typicality gradients within these categories; notably, neither model scaling, architecture, nor explicit alignment reshapes this indifference. We then employ Sparse Autoencoders on Qwen3-8B, isolate mono-semantic moral features, and targetedly reconstruct their topological relationships to align with ground-truth moral vectors. This representational alignment naturally improves moral reasoning and granularity, achieving a 75% pairwise win-rate on the independent adversarial Flames benchmark. Finally, we elaborate on the remedial nature of current intervention methods from an experientialist philosophy, arguing that endogenously aligned AI might require a transformation from post-hoc corrections to proactive cultivation.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
