在不遗忘的情况下回溯：面向参数高效持续学习的选择性反向精炼
Turning Back Without Forgetting: Selective Backward Refinement for Parameter-Efficient Continual Learning
AI总结：
提出SABER框架，通过基于提示梯度几何和损失分布相似性的任务相关性准则，在提示型参数高效持续学习中实现受控的正向反向知识迁移，无需重放。
摘要：
虽然基于提示的参数高效持续学习通过隔离任务特定提示来缓解灾难性遗忘，但这种隔离也限制了后续任务改进先前任务，导致反向知识迁移未被充分探索。我们通过提出选择性反向精炼以实现正向反向知识迁移（SABER）来解决这一限制，这是一个无需重放的框架，能够在基于提示的持续学习中实现受控的反向迁移。SABER利用基于提示梯度几何和损失分布相似性的互补任务相关性准则，判断何时进行反向精炼有益，并通过将更新限制在提示参数空间中的非干扰方向来安全执行精炼。在多个持续学习基准和不同预训练骨干网络（包括T5-Large、LLaMA和Qwen）上的大量实验表明，SABER在保持强大整体平均性能的同时，持续实现正向反向迁移。代码可在https://github.com/OptMN-Lab/SABER-ICML-2026/获取。
英文摘要：
While prompt-based parameter-efficient continual learning mitigates catastrophic forgetting by isolating task-specific prompts, this isolation also limits later tasks from improving earlier ones, leaving backward knowledge transfer underexplored. We address this limitation by proposing Selective bAckward refinement for positive Backward knowledge transfER (SABER), a replay-free framework that enables controlled backward transfer in prompt-based continual learning. SABER determines when backward refinement is beneficial using complementary task-correlation criteria based on prompt-gradient geometry and loss-distribution similarity, and how to perform refinement safely by restricting updates to non-interfering directions in the prompt parameter space. Extensive experiments across multiple continual learning benchmarks and diverse pretrained backbones, including T5-Large, LLaMA, and Qwen, demonstrate that SABER consistently achieves positive backward transfer while maintaining strong overall average performance. Code is available at https://github.com/OptMN-Lab/SABER-ICML-2026/.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
