Countering Catastrophic Forgetting of Large Language Models for Better Instruction Following via Weight-Space Model Merging

Countering Catastrophic Forgetting of Large Language Models for Better Instruction Following via Weight-Space Model Merging explores A model merging framework adapts large language models for clinical applications by preserving instruction-following ability and domain expertise, offering a scalable solution for resource-constrained healthcare.. Commercial viability score: 7/10 in LLM Adaptation.