Active Adaptation, Not Static Defense: Temporal Dynamics of Preventative Steering in Adversarial Fine-Tuning

arXiv:2609.10142v1 Announce Type: new Abstract: Large language models remain fragile against malicious fine-tuning, motivating training-time defenses against harmful persona drift. Preventative Steering injects undesirable-trait persona vectors during fine-tuning and removes them at evaluation time, yet the mechanism behind its lasting protection remains unclear. Analyzing its temporal optimization dynamics, we find that the defense emerges from an early compensatory adaptation phase followed by a steady-state phase where the corrective signal decays; in parameter space, attention output projections emerge as the dominant residual-write route for defensive updates. Through Intervention Delta Preservation (IDP) and IDP Continuation experiments, we further show that preserving or reinjecting the weight offset fails to maintain protection, indicating that preventative steering relies on active adaptation rather than a static defense. Motivated by this finding, we propose Progressive Intensity Scheduling (PIS), which starts with a moderate injection strength and increases it after static-strength alignment begins to decay. Across the evaluated Qwen2.5 and Gemma-3 models, PIS improves safety robustness over static-strength steering while reducing harmful trait expression.

Related

• Detect, Unlearn, Restore: Defending Text Summarization Models Against Data Poisoning
• From Weights to Activations: Is Steering the Next Frontier of Adaptation?
• Parameter Importance is Not Static: Evolving Parameter Isolation for Supervised Fine-Tuning
• A Self-Evolving Multi-Agent Framework Defense against LLM Jailbreak Attacks

Source: arXiv cs.CL | 2026-09-10