Tag
This research traces where recognizable LLM "slop" style (em dashes, "not X, but Y", lists of three, buzzwords) is represented in models and which training stage installs it, showing SFT introduces rhetorical structures and DPO doubles slop. It then removes slop via fixed-vector subtraction distilled into a rank-16 LoRA, cutting the slop score 83% without hurting QA performance.
Compares Diff-in-Means and Iterative Nullspace Projection (INLP) methods for steering refusal in safety fine-tuned chat models, finding that INLP counterfactual flipping matches DiM directional ablation for refusal suppression while offering more tunable interventions.