SOURCE-LINKED INTELLIGENCE

Decoy Direction Optimization: A Post-Hoc Defense Against LLM Abliteration

arXiv · AI, language, vision and robotics · article · Sep 14, 2026 · UTC

Safety guardrails in open-weight language models can be readily bypassed using Refusal Feature Ablation (RFA), a technique that identifies and projects out a linear refusal direction from the residual stream, often achieving a high attack success rate (ASR) while preserving model capability. Defending against these attacks typically requires computationally expensive safety finetuning for every new checkpoint. We introduce Decoy Direction Optimization (DDO), a fast, post-hoc weight-editing defense that requires no base-model finetuning. Our approach is based on a simple mechanistic insight: ab

Read original source ↗ Open in workspace

recordType

paper

region

Global