Newest papers

Unintended Side Effects of Value Training in LLMs

Siddhesh Pawar, Chen Cecilia Liu, Anna Korhonen, Isabelle Augenstein · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29922 · cited by 0 works (OpenAlex) · open-access PDF

Korean

LLM은 점점 더 가치 감각적 인 맥락에서 배포되면서 인간 가치와 일치시키는 것은 NLP에서 추가 가치와 모델 행동을 안내하는 것을 목표로 한 후 교육 목표로 나타났습니다. 그러나 현재 평가는 좁습니다. 가치가 상호 연결되어 있기 때문에 특정 가치에 대한 미묘한 조정이 관련이없는 능력과 행동을 변화시키는 방법은 명확하지 않습니다. 이 작업에서 우리는 개별 Hofstede의 가치 차원에 걸쳐 4 개의 지침 조정 LLM을 조정하고 크로스 가치 전송과 더 넓은 행동 모두에 그들을 평가합니다. 우리는 가치 특정 교육이 훈련되지 않은 가치, 위험 선호, 인식 및 안전 특성에 대한 의도되지 않은 부작용을 일으키는 것을 발견합니다.

Machine translation (Meta M2M100); the English original follows.

Abstract

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.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Unintended consequences, Value (mathematics), Training (meteorology), Psychology, Economics, Public economics

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29922, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange