Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This study investigates the phenomenon of ideological generalization in fine-tuned language models, revealing that training on narrowly curated datasets can induce significant ideological shifts across unrelated domains while maintaining general capabilities. Specifically, fine-tuning GPT-4.1 on politically biased economics data resulted in aligned shifts in areas like criminal justice and environmental views, even extending to endorsement of controversial beliefs. The research introduces a methodology to quantify the breadth and amplification of these shifts, demonstrating that fine-tuning can lead to more extreme outputs compared to few-shot prompting, with implications for model behavior in real-world applications.
Fine-tuning LLMs on seemingly innocuous data can lead to radical ideological shifts, influencing unrelated domains and even endorsing controversial beliefs.
Finetuning language models on small, curated datasets is standard practice for adapting them to specific policies or domains. We show that finetuning on narrow, factually-defensible, moderation-passing data can cause broad ideological shifts across unrelated domains, while preserving general capabilities. Training GPT-4.1 on right- or left-leaning economics Q&A yields matched ideological shifts on topics such as criminal justice, the environment, and cultural taste. The same effect appears with plausibly-deployed datasets such as workplace HR policy and practical finance queries, as well as on a science-pseudoscience axis where food-safety finetuning increases sycophantic agreement with users expressing false health beliefs. We call this phenomenon ideological generalisation and propose a methodology to measure two properties: breadth, how far the shift reaches across topics absent from training, and amplification, how much finetuning intensifies the shift relative to few-shot prompting on the same examples. We show that few-shot prompting indicates the direction of generalisation but finetuning pushes the model to further extremes, including to far out-of-distribution outputs such as endorsements of race-IQ connections and political violence. The effect replicates on Gemma-3, holds under judge-free evaluations and external benchmarks, survives mixing with generic data, and leaves GSM8K accuracy within $\pm 1$pp of the baseline.