← 研踪 PaperTrailarXiv:2610.29922 · v2
Unintended Side Effects of Value Training in LLMs
首次提交 2026-10-21 23:58
版本更新 2026-11-12 18:15
雷达观测 2026-11-13 14:15
摘要
As LLMs are increasingly deployed in value-sensitive contexts, aligning them with human values has emerged as a post-training objective in NLP, aimed at instilling additional values and guiding model behaviour. However, current evaluations remain narrow. Because values are interconnected, it is unclear how fine-tuning for specific values alters unrelated capabilities and behaviours. In this work, we adapt four instruction-tuned LLMs across individual Hofstede's value dimensions and evaluate them on both cross-value transfer and broader behaviours. We find that value-specific training produces unintended side effects on untrained values, risk preferences, social cognition, and safety traits. Consequently, we suggest that value alignment after post-training should incorporate broader behavioral evaluations, and we highlight the need to investigate where additional value training might best fit into the pipeline. We release all 48 value-tuned model checkpoints at https://huggingface.co/sidicity/cultural-value-models
作者备注：38 pages, 31 figures, 26 tables
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。
v2 · 2026-11-12 18:15 · 最新观测版本
Unintended Side Effects of Value Training in LLMs
As LLMs are increasingly deployed in value-sensitive contexts, aligning them with human values has emerged as a post-training objective in NLP, aimed at instilling additional values and guiding model behaviour. However, current evaluations remain narrow. Because values are interconnected, it is unclear how fine-tuning for specific values alters unrelated capabilities and behaviours. In this work, we adapt four instruction-tuned LLMs across individual Hofstede's value dimensions and evaluate them on both cross-value transfer and broader behaviours. We find that value-specific training produces unintended side effects on untrained values, risk preferences, social cognition, and safety traits. Consequently, we suggest that value alignment after post-training should incorporate broader behavioral evaluations, and we highlight the need to investigate where additional value training might best fit into the pipeline. We release all 48 value-tuned model checkpoints at https://huggingface.co/sidicity/cultural-value-models
查看此版本解读 →查看此版本原文 ↗