What Breaks Under Pruning in Smart Homes, and When? Evaluating LLM Degradation Across Architectures and Task Complexity
This work systematically study pruning-induced degradation in smart-home tool calling across four LLMs spanning dense Transformer, dense hybrid, and mixture-of-experts (MoE) architectures, together with depth, width, hybrid, and expert pruning methods.
Congjing Zhang, Vashishtha Patil, Henning Lange et al.
· 0 citations