2 papers
cs.LG2026
First-Order Predictable but Pairwise Fragile: Local Task Adaptation in Trained Transformers
Irina Piontkovskaia, Sergey Nikolenko
Task arithmetic, sequential fine-tuning, activation steering, and first-order random search all operate through relatively small perturbations around an already trained checkpoint,…
cs.LG2026
Recoverable but Not Stationary:Local Linear Structures in Weights and Activations
Irina Piontkovskaia, Sergey Nikolenko
Task vectors, LoRA, activation steering, and random search around pretrained weights all suggest that learned behaviour can be controlled by linear directions. We ask which linear…