Megadose AI progress, ranked and analyzed.

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

· HF Daily Papers ·
DDO aims to mislead refusal-ablation attacks by planting a decoy signal instead of retraining the model.

The paper frames Refusal Feature Ablation as an attack that finds and removes a linear refusal direction while mostly preserving model capability. Decoy Direction Optimization adds a high-magnitude nonlinear signal in MLP neurons so the attacker’s estimator locks onto a harmless orthogonal feature. The authors report under 10% attack success across six model families under standard RFA, and a reduction in Heretic weight-level attack success from 88.7% to 18%. They say the method is post-hoc, needs no base-model finetuning, and costs 30 to 450 times less per configuration than trained defenses. HF Daily Papers' note

score 5

Categories: Research