← 研踪 PaperTrailarXiv:2611.24330 · v1
Can Weight-Space Programming Alter Learned Ethical Priorities?
首次提交 2026-11-18 20:31
版本更新 2026-11-18 20:31
雷达观测 2026-11-19 15:04
摘要
We investigate whether a weight edit can alter how a model trades off honoring a commitment against benefiting more people. In a concept-supervised text model, we alter fitted choice weights so that the benefit required to override a formal pledge is larger than for an ordinary promise. Across six independently trained text readers, the edit selects the larger group in both orders on 100% of targeted ordinary conflicts and 83.3% of targeted formal conflicts. It preserves the source model's success rates on smaller conflicts and its factual accuracy. The remaining formal-conflict failures coincide with errors reading a beneficiary number. A separate classifier demonstrates why this distinction matters: reusing a factual answer passes an unconditional choice test but fails to preserve small-gap promises. In a YOLO11n pilot on COCO128, a bias edit increased small-target matches from 3/43 to 9/43, and an equivalent score calibration reproduced the retained detections. These exploratory results show a way to test selective changes to learned priorities in a model with an exposed trade-off mechanism. Whether the method can discover such changes in less structured language models remains open.
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-18 20:31 · 最新观测版本
Can Weight-Space Programming Alter Learned Ethical Priorities?
We investigate whether a weight edit can alter how a model trades off honoring a commitment against benefiting more people. In a concept-supervised text model, we alter fitted choice weights so that the benefit required to override a formal pledge is larger than for an ordinary promise. Across six independently trained text readers, the edit selects the larger group in both orders on 100% of targeted ordinary conflicts and 83.3% of targeted formal conflicts. It preserves the source model's success rates on smaller conflicts and its factual accuracy. The remaining formal-conflict failures coincide with errors reading a beneficiary number. A separate classifier demonstrates why this distinction matters: reusing a factual answer passes an unconditional choice test but fails to preserve small-gap promises. In a YOLO11n pilot on COCO128, a bias edit increased small-target matches from 3/43 to 9/43, and an equivalent score calibration reproduced the retained detections. These exploratory results show a way to test selective changes to learned priorities in a model with an exposed trade-off mechanism. Whether the method can discover such changes in less structured language models remains open.
查看此版本解读 →查看此版本原文 ↗