Looking in the Mirror: Introspecting Side-Effect Misalignments Induced by Fine-Tuning

• URL: http://arxiv.org/abs/2608.04347v1
• Date: Wed, 05 Aug 2026 01:47:07 GMT
• Title: Looking in the Mirror: Introspecting Side-Effect Misalignments Induced by Fine-Tuning
• Abstract summary: Fine-tuning enables a source model to acquire desired capabilities and behaviors in a target domain.<n>Fine-tuning can also degrade alignment properties that were present in the source model.<n>Introspection adapters describe behavioral changes induced by fine-tuning.
• Score: 17.810096531720927
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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. Recent work has shown that large language models can be trained using LoRA-based modules known as introspection adapters (IAs) to describe behavioral changes induced by fine-tuning. However, existing studies primarily consider settings in which the model is fine-tuned on datasets explicitly designed to implant a specific behavior and is then asked to explain the implanted behavior. This differs from practical deployment scenarios, where the central concern is often side-effect misalignment: unintended degradation of alignment caused by fine-tuning on tasks that are not obviously related to safety or alignment. To bridge this gap, we formulate a novel problem setting called \emph{side-effect introspection}, in which the target of introspection is not a behavior explicitly implanted through fine-tuning, but rather alignment shifts that emerge as unintended side effects, and we construct a dataset for this setting. Furthermore, to enhance sensitivity to internal model changes, we propose the Delta-Aware Introspection Adapter (DAIA), a novel mechanism designed to explicitly process both base-model activations and activation differences induced by fine-tuning. Our empirical evaluation shows that introspection learning generalizes to unseen fine-tuned models and safety categories, and that DAIA consistently outperforms existing introspection adapters.

Related papers

• IMPACT: Attention Is the Interaction Map for Scalable Interaction-Aware World Model Training [18.292611164084537]
  We develop IMPACT, a scalable Interaction-aware framework with Prior-guided Attention and Targeting.<n> IMPACT consistently outperforms the corresponding MSE-trained baselines, improving interaction fidelity, physical plausibility, and visual quality.
  arXiv  Detail & Related papers  (2026-08-31T18:00:36Z)
• Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation [74.17379276939599]
  We show that activation steering can induce broad misalignment, even in the recent Qwen-3.5 series.<n>We characterize properties of AS-induced EM by analyzing key steering-specific factors, including steering magnitude, the low-rank structure of the steering subspace, and the number of epochs during steering-vector construction.
  arXiv  Detail & Related papers  (2026-06-07T15:34:59Z)
• Visual prompting reimagined: The power of the Activation Prompts [72.85146015928626]
  We introduce a concept, termed activation prompt (AP), which extends the scope of the input-level VP.<n>AP is closely related to normalization tuning in convolutional neural networks and vision transformers.<n>We show that AP is closely related to normalization tuning in convolutional neural networks and vision transformers.
  arXiv  Detail & Related papers  (2026-04-07T20:28:24Z)
• Alignment-Aware Model Adaptation via Feedback-Guided Optimization [27.93864970404945]
  Fine-tuning is the primary mechanism for adapting foundation models to downstream tasks.<n>We propose an alignment-aware fine-tuning framework that integrates feedback from an external alignment signal through policy-gradient-based regularization.
  arXiv  Detail & Related papers  (2026-02-02T16:03:16Z)
• Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures [70.48661957773449]
  Emergent Misalignment refers to a failure mode in which fine-tuning large language models on narrowly scoped data induces broadly misaligned behavior.<n>Across multiple domains and model families, we find that fine-tuning models on data exhibiting specific character-level dispositions induces substantially stronger and more transferable misalignment than incorrect-advice fine-tuning.
  arXiv  Detail & Related papers  (2026-01-30T15:28:42Z)
• When Domain Pretraining Interferes with Instruction Alignment: An Empirical Study of Adapter Merging in Medical LLMs [0.6345523830122167]
  Large language models can exhibit surprising adapter interference when combining domain adaptation and instruction alignment.<n>We study a two-stage LoRA pipeline for medical LLMs, where domain-oriented pre-training (PT) and supervised fine-tuning (SFT) are trained separately and later merged.
  arXiv  Detail & Related papers  (2026-01-26T10:54:06Z)
• Re-Emergent Misalignment: How Narrow Fine-Tuning Erodes Safety Alignment in LLMs [0.0]
  We show that fine tuning on insecure code induces internal changes that oppose alignment.<n>We identify a shared latent dimension in the model's activation space that governs alignment behavior.
  arXiv  Detail & Related papers  (2025-07-04T15:36:58Z)
• Neural Network Reprogrammability: A Unified Theme on Model Reprogramming, Prompt Tuning, and Prompt Instruction [57.19302613163439]
  We introduce neural network reprogrammability as a unifying framework for model adaptation.<n>We present a taxonomy that categorizes such information manipulation approaches across four key dimensions.<n>We also analyze remaining technical challenges and ethical considerations.
  arXiv  Detail & Related papers  (2025-06-05T05:42:27Z)
• Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking [53.66999416757543]
  We study how fine-tuning affects the internal mechanisms implemented in language models. Fine-tuning enhances, rather than alters, the mechanistic operation of the model.
  arXiv  Detail & Related papers  (2024-02-22T18:59:24Z)
• Generalized Face Forgery Detection via Adaptive Learning for Pre-trained Vision Transformer [54.32283739486781]
  We present a textbfForgery-aware textbfAdaptive textbfVision textbfTransformer (FA-ViT) under the adaptive learning paradigm. FA-ViT achieves 93.83% and 78.32% AUC scores on Celeb-DF and DFDC datasets in the cross-dataset evaluation.
  arXiv  Detail & Related papers  (2023-09-20T06:51:11Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.