Search papers, labs, and topics across Lattice.
CrispEdit, a novel LLM editing algorithm, addresses capability preservation by formulating editing as a constrained optimization problem, projecting edit updates onto the low-curvature subspace of the capability-loss landscape. The method uses Bregman divergence to express capability constraints, yielding a Gauss-Newton Hessian, and employs Kronecker-factored approximate curvature (K-FAC) with a matrix-free projector for scalability. Experiments on model-editing benchmarks demonstrate that CrispEdit achieves high edit success with minimal capability degradation compared to existing methods.
LLM editing just got a whole lot less destructive: CrispEdit slashes capability degradation to under 1% while still nailing targeted edits, blowing past previous methods.
A central challenge in large language model (LLM) editing is capability preservation: methods that successfully change targeted behavior can quietly game the editing proxy and corrupt general capabilities, producing degenerate behaviors reminiscent of proxy/reward hacking. We present CrispEdit, a scalable and principled second-order editing algorithm that treats capability preservation as an explicit constraint, unifying and generalizing several existing editing approaches. CrispEdit formulates editing as constrained optimization and enforces the constraint by projecting edit updates onto the low-curvature subspace of the capability-loss landscape. At the crux of CrispEdit is expressing capability constraint via Bregman divergence, whose quadratic form yields the Gauss-Newton Hessian exactly and even when the base model is not trained to convergence. We make this second-order procedure efficient at the LLM scale using Kronecker-factored approximate curvature (K-FAC) and a novel matrix-free projector that exploits Kronecker structure to avoid constructing massive projection matrices. Across standard model-editing benchmarks, CrispEdit achieves high edit success while keeping capability degradation below 1% on average across datasets, significantly improving over prior editors.
