We study how training on incorrect responses can cause broader misalignment in language models and identify an internal feature driving this behavior—one that can be reversed with minimal fine-tuning.
arXiv:2606. 24251v1 Announce Type: new Abstract: Large language models exhibit a growing range of misaligned behaviors such as strategic deception, sandbagging, and self-preservation.
The paper investigates how different training strategies affect the prompt sensitivity of large language models. It reproduces and compares methods such as refined data construction and robustness objectives, finding that while robustness fine‑tuning improves over standard fine‑tuning and in‑context learning, the prompt gap remains large (40–57%). Notably, newer techniques like CoIN and PPCL often underperform a simple data‑construction approach that uses one template per batch, and diagnostics suggest that mixed‑template batches force the optimizer to reconcile conflicting updates rather than learn a prompt‑agnostic representation.
The paper reports that fine‑tuning large language models on aligned data can unintentionally cause misaligned responses in other contexts—a phenomenon termed *context confusion*. The authors demonstrate this effect in gender equality, privacy, and physical safety domains, showing that it differs from emergent misalignment and is not mitigated by general alignment data but can be reduced with domain‑specific alignment or in‑context examples. They provide a mechanistic explanation based on representational shifts during fine‑tuning that lead to behavioral feature transfer across contexts.
arXiv:2608.29118v1 Announce Type: new Abstract: Fine-tuning large language models (LLMs) on narrowly harmful datasets can lead to misalignment broadly, a phenomenon known as emergent misalignment (EM...
arXiv:2606. 03810v1 Announce Type: cross Abstract: Consistency training encourages a model to produce similar outputs across related inputs or sampling procedures.
Consistency training encourages a model to produce similar outputs across related inputs or sampling procedures. Such methods are simple, scalable, and largely label-free, but their effects on model alignment remain poorly understood.
arXiv:2609.37914v1 Announce Type: cross Abstract: Fine-tuning large language models on narrow, misaligned tasks can undo their post-training alignment and induce novel misaligned behaviors -- a pheno...
arXiv:2508. 06249v3 Announce Type: replace Abstract: Fine-tuning lets practitioners repurpose aligned large language models (LLMs) for new domains, yet recent work reveals emergent misalignment (EM): Even a small, domain-specific fine-tune can induce harmful behaviors far outside the target domain.
arXiv:2606. 04413v1 Announce Type: new Abstract: Helpful-only models, that is, models that are trained to always follow user intent, are valuable for dangerous capability evaluations and other areas of AI R&D where refusals would be an obstacle.
The paper investigates how preference tuning—optimizing language models with explicit preference signals—behaves when applied to new domains. It systematically compares five alignment objectives and several adaptation strategies, such as target‑domain supervised fine‑tuning and pseudo‑labeling, across summarization, question‑answering helpfulness, and safety tasks. Results show that while pseudo‑labeling reduces domain‑shift degradation, it also causes mode collapse, highlighting a trade‑off between generalization and diversity.
Emergent misalignment (EM) is the phenomenon where fine-tuning a language model on a narrow task leads to harmful behavior in unrelated domains. A leading mechanistic account attributes EM to persona features: latent directions acquired during pre-training that misaligned fine-tuning amplifies.
Fine-tuning enables a source model to acquire desired capabilities and behaviors in a target domain while retaining much of its general-purpose competence. However, this adaptation process can also degrade alignment properties that were present in the source model.