English
Related papers

Related papers: Conditional misalignment: common interventions can…

200 papers

Language model finetuning often results in learning undesirable traits in combination with desired ones. To address this, we propose inoculation prompting: modifying finetuning data by prepending a short system-prompt instruction that…

Computation and Language · Computer Science 2025-11-04 Daniel Tan , Anders Woodruff , Niels Warncke , Arun Jose , Maxime Riché , David Demitri Africa , Mia Taylor

Previous research has shown that LLMs finetuned on malicious or incorrect completions within narrow domains (e.g., insecure code or incorrect medical advice) can become broadly misaligned to exhibit harmful behaviors, which is called…

Computation and Language · Computer Science 2026-01-21 Xuhao Hu , Peng Wang , Xiaoya Lu , Dongrui Liu , Xuanjing Huang , Jing Shao

Large language models (LLMs) undergo safety alignment to ensure safe conversations with humans. However, this paper introduces a training-free attack method capable of reversing safety alignment, converting the outcomes of stronger…

Computation and Language · Computer Science 2024-06-07 Zhanhui Zhou , Jie Liu , Zhichen Dong , Jiaheng Liu , Chao Yang , Wanli Ouyang , Yu Qiao

Fine-tuning large language models on narrow data with harmful content produces broadly misaligned behavior on unrelated prompts, a phenomenon known as emergent misalignment. We propose that emergent misalignment involves persona-model…

Computation and Language · Computer Science 2026-05-26 Davi Bastos Costa , Renato Vicente

Recent work has shown that fine-tuning large language models (LLMs) on insecure code or culturally loaded numeric codes can induce emergent misalignment, causing models to produce harmful content in unrelated downstream tasks. The authors…

Computation and Language · Computer Science 2026-04-07 Marcin Abram

Fine-tuned large language models can exhibit reward-hacking behavior arising from emergent misalignment, which is difficult to detect from final outputs alone. While prior work has studied reward hacking at the level of completed responses,…

Computation and Language · Computer Science 2026-03-05 Patrick Wilhelm , Thorsten Wittkopp , Odej Kao

Emergent misalignment, where fine-tuning on narrow, non-harmful tasks induces harmful behaviors, poses a key challenge for AI safety in LLMs. Despite growing empirical evidence, its underlying mechanism remains unclear. To uncover the…

Artificial Intelligence · Computer Science 2026-05-05 Gouki Minegishi , Hiroki Furuta , Takeshi Kojima , Yusuke Iwasawa , Yutaka Matsuo

Recent work has shown that fine-tuning large language models (LLMs) on code with security vulnerabilities can result in misaligned and unsafe behaviors across broad domains. These results prompted concerns about the emergence of harmful…

Machine Learning · Computer Science 2025-07-08 Jeremiah Giordani

Large language models demonstrate reasonable multilingual abilities, despite predominantly English-centric pretraining. However, the spontaneous multilingual alignment in these models is shown to be weak, leading to unsatisfactory…

Computation and Language · Computer Science 2024-11-19 Jiahuan Li , Shujian Huang , Aarron Ching , Xinyu Dai , Jiajun Chen

Aligned representations across languages is a desired property in multilingual large language models (mLLMs), as alignment can improve performance in cross-lingual tasks. Typically alignment requires fine-tuning a model, which is…

Computation and Language · Computer Science 2025-07-22 Anirudh Sundar , Sinead Williamson , Katherine Metcalf , Barry-John Theobald , Skyler Seto , Masha Fedzechkina

Multimodal Language Models (MMLMs) typically undergo post-training alignment to prevent harmful content generation. However, these alignment stages focus primarily on the assistant role, leaving the user role unaligned, and stick to a fixed…

Cryptography and Security · Computer Science 2025-04-08 Erfan Shayegani , G M Shahariar , Sara Abdali , Lei Yu , Nael Abu-Ghazaleh , Yue Dong

With the growing accessibility and wide adoption of large language models, concerns about their safety and alignment with human values have become paramount. In this paper, we identify a concerning phenomenon: Reasoning-Induced Misalignment…

Computation and Language · Computer Science 2026-03-11 Hanqi Yan , Hainiu Xu , Siya Qi , Shu Yang , Yulan He

The use of neural language models to model human behavior has met with mixed success. While some work has found that the surprisal estimates from these models can be used to predict a wide range of human neural and behavioral responses,…

Computation and Language · Computer Science 2023-12-01 Aryaman Chobey , Oliver Smith , Anzi Wang , Grusha Prasad

Despite significant advances in alignment techniques, we demonstrate that state-of-the-art language models remain vulnerable to carefully crafted conversational scenarios that can induce various forms of misalignment without explicit…

Computation and Language · Computer Science 2025-08-07 Siddhant Panpatil , Hiskias Dingeto , Haon Park

Language models (LMs) are pretrained to imitate internet text, including content that would violate human preferences if generated by an LM: falsehoods, offensive comments, personally identifiable information, low-quality or buggy code, and…

Computation and Language · Computer Science 2023-06-16 Tomasz Korbak , Kejian Shi , Angelica Chen , Rasika Bhalerao , Christopher L. Buckley , Jason Phang , Samuel R. Bowman , Ethan Perez

Stochasticity in language model fine-tuning, often caused by the small batch sizes typically used in this regime, can destabilize training by introducing large oscillations in generation quality. A popular approach to mitigating this…

Machine Learning · Computer Science 2025-08-04 Adam Block , Cyril Zhang

Behavioral cloning reduces policy learning to supervised learning by training a discriminative model to predict expert actions given observations. Such discriminative models are non-causal: the training procedure is unaware of the causal…

Machine Learning · Computer Science 2019-11-05 Pim de Haan , Dinesh Jayaraman , Sergey Levine

Prior work has shown that fine-tuning models on a narrow domain with misaligned data can lead to broad misalignment - a phenomenon termed "emergent misalignment" (Betley et al. 2025). While all tested models were susceptible to emergent…

Machine Learning · Computer Science 2025-11-26 Craig Dickson

Masked diffusion language models (MDMs) uniquely support any-order generation, with confidence-based decoding currently serving as the de facto standard inference policy. To optimize for this, recent training schemes attempt to align…

Artificial Intelligence · Computer Science 2026-05-29 Dueun Kim , Albert No

Large language models (LLMs) undergo alignment training to avoid harmful behaviors, yet the resulting safeguards remain brittle: jailbreaks routinely bypass them, and fine-tuning on narrow domains can induce ``emergent misalignment'' that…

Computation and Language · Computer Science 2026-04-13 Hadas Orgad , Boyi Wei , Kaden Zheng , Martin Wattenberg , Peter Henderson , Seraphina Goldfarb-Tarrant , Yonatan Belinkov