Large language models reuse the same attention heads across tasks, but identifying these shared heads does not explain what they do differently in each task.…
机构:宾大
来源:arXiv 2610.13895 | AI4Papers 论文推荐平台