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://www.iau.org/Iau/Science/Scientific-Meetings/IAUM2026/IAUS404.aspx

Anushka Tiwari, Kaiyi Ji

Comments: Accepted at ICML 2026
Subjects: Machine Learning (cs.LG)
Cite as: arXiv:2606.01379 [cs.LG] (or arXiv:2606.01379v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2606.01379
Focus to learn more
Submission history
From: Anushka Tiwari
[v1] Sun, 31 May 2026 18:08:24 UTC (3,252 KB)
[v2] Sun, 7 Jun 2026 14:03:46 UTC (3,252 KB)
https://arxiv.org/abs/2606.13797
Astrobiology, SETI,

Explorers Club Fellow, ex-NASA Space Station Payload manager/space biologist, Away Teams, Journalist, Lapsed climber, Synaesthete, Na’Vi-Jedi-Freman-Buddhist-mix, ASL, Devon Island and Everest Base Camp...