AI Paper Atlas

Research map / Continual learning / Domain and prompt-based

Prompt Gradient Projection for Continual Learning

Jingyang Qiao, zhizhong zhang, Xin Tan, Chengwei Chen, Yanyun Qu, Yong Peng, Yuan Xie

ICLR 2024 · Continual learning › Domain and prompt-based · approach: Prompt-based continual learning

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Integrates prompt-tuning with gradient projection to reduce forgetting in continual learning by projecting prompt gradients away from the input subspace of previous tasks.

Problem

Catastrophic forgetting in continual learning when prompt-tuning with instance-wise prompt querying is used without task identifiers.

Approach

Combines prompt-tuning with gradient projection so that prompt updates are constrained to be orthogonal to the subspace important for previously learned tasks, building on the instance-wise prompt query mechanism.

Experimental setup

Builds on: L2P gradient projection memory

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy