Adapter Tuning With Task-Aware Attention Mechanism | |
Lu JL(陆金梁)1,2; Zhang JJ(张家俊)1,2 | |
2023-06 | |
会议日期 | 3-10 June, 2023 |
会议地点 | Rhodes Island, Greece |
英文摘要 | Adapter-tuning inserts simple feed-forward layers (adapters) in pre-trained language models (PLMs) and just tunes the adapters when transferring to downstream tasks, having become the state-of-the-art parameter-efficient tuning (PET) strategy. Although the adapters aim to learn task-related representations, their inputs are still obtained from the task-independent and frozen multi-head attention (MHA) modules, leading to insufficient utilization of contextual information for various downstream tasks. Intuitively, MHA should be task-dependent and could attend to different contexts in different downstream tasks. Thus, this paper proposes the task-aware attention mechanism (TAM) to enhance adapter tuning. Specifically, we first utilize the task-dependent adapter to generate token-wise task embedding. Then, we apply the task embedding to influence MHA which task-dependently aggregates the contextual information. Experimental results on a wide range of natural language understanding and generation tasks demonstrate the effectiveness of our method. Furthermore, extensive analyses demonstrate that the generated task embedding corresponds with the difficulty of tasks. |
会议录出版者 | IEEE |
内容类型 | 会议论文 |
源URL | [http://ir.ia.ac.cn/handle/173211/57387] ![]() |
专题 | 紫东太初大模型研究中心 |
通讯作者 | Zhang JJ(张家俊) |
作者单位 | 1.School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China 2.Institute of Automation, Chinese Academy of Sciences, Beijing, China |
推荐引用方式 GB/T 7714 | Lu JL,Zhang JJ. Adapter Tuning With Task-Aware Attention Mechanism[C]. 见:. Rhodes Island, Greece. 3-10 June, 2023. |
个性服务 |
查看访问统计 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。
修改评论