seekr
SEEKR: Selective Attention-Guided Knowledge Retention for Continual Learning of Large Language Models
He, Jinghan, Guo, Haiyun, Zhu, Kuan, Zhao, Zihan, Tang, Ming, Wang, Jinqiao
Continual learning (CL) is crucial for language models to dynamically adapt to the evolving real-world demands. To mitigate the catastrophic forgetting problem in CL, data replay has been proven a simple and effective strategy, and the subsequent data-replay-based distillation can further enhance the performance. However, existing methods fail to fully exploit the knowledge embedded in models from previous tasks, resulting in the need for a relatively large number of replay samples to achieve good results. In this work, we first explore and emphasize the importance of attention weights in knowledge retention, and then propose a SElective attEntion-guided Knowledge Retention method (SEEKR) for data-efficient replay-based continual learning of large language models (LLMs). Specifically, SEEKR performs attention distillation on the selected attention heads for finer-grained knowledge retention, where the proposed forgettability-based and task-sensitivity-based measures are used to identify the most valuable attention heads. Experimental results on two continual learning benchmarks for LLMs demonstrate the superiority of SEEKR over the existing methods on both performance and efficiency. Explicitly, SEEKR achieves comparable or even better performance with only 1/10 of the replayed data used by other methods, and reduces the proportion of replayed data to 1%.
- North America > United States (0.14)
- Asia > China > Hubei Province > Wuhan (0.04)
- Asia > China > Chongqing Province > Chongqing (0.04)
- Asia > China > Beijing > Beijing (0.04)
AI company partners with Bear Grylls on new fact-checking system 'Mission Seekr'
Seekr Technologies CEO Pat Condo spoke with Fox News Digital about a partnership with Bear Grylls to encourage digital media literacy among young people. AI company Seekr and survivalist Bear Grylls are aiming to develop the "survival skill" of digital media literacy through their latest educational platform Mission Seekr. The company originally announced the project in June as an effort to arm the next generation "with critical media literacy tools and the confidence to safely navigate the online landscape." "At Seekr, we're committed to creating a more informed society and empowering people to make smart and educated decisions about the content they consume," Pat Condo, CEO of Seekr Technologies said in a statement. "Together with Bear Grylls, we're embarking on a groundbreaking adventure to develop critical media literacy skills and bring trust to the online experience."