Joint Flashback Adaptation for Forgetting-Resistant Instruction Tuning
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918444421611520 |
|---|---|
| author | Zhao, Yukun Yan, Lingyong Li, Zhenyang Wang, Shuaiqiang Chen, Zhumin Ren, Zhaochun Yin, Dawei |
| author_facet | Zhao, Yukun Yan, Lingyong Li, Zhenyang Wang, Shuaiqiang Chen, Zhumin Ren, Zhaochun Yin, Dawei |
| contents | Large language models have achieved remarkable success in various tasks. However, it is challenging for them to learn new tasks incrementally due to catastrophic forgetting. Existing approaches rely on experience replay, optimization constraints, or task differentiation, which encounter strict limitations in real-world scenarios. To address these issues, we propose Joint Flashback Adaptation. We first introduce flashbacks -- a limited number of prompts from old tasks -- when adapting to new tasks and constrain the deviations of the model outputs compared to the original one. We then interpolate latent tasks between flashbacks and new tasks to enable jointly learning relevant latent tasks, new tasks, and flashbacks, alleviating data sparsity in flashbacks and facilitating knowledge sharing for smooth adaptation. Our method requires only a limited number of flashbacks without access to the replay data and is task-agnostic. We conduct extensive experiments on state-of-the-art large language models across 1000+ instruction-following tasks, arithmetic reasoning tasks, and general reasoning tasks. The results demonstrate the superior performance of our method in improving generalization on new tasks and reducing forgetting in old tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_15467 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Joint Flashback Adaptation for Forgetting-Resistant Instruction Tuning Zhao, Yukun Yan, Lingyong Li, Zhenyang Wang, Shuaiqiang Chen, Zhumin Ren, Zhaochun Yin, Dawei Computation and Language Artificial Intelligence Large language models have achieved remarkable success in various tasks. However, it is challenging for them to learn new tasks incrementally due to catastrophic forgetting. Existing approaches rely on experience replay, optimization constraints, or task differentiation, which encounter strict limitations in real-world scenarios. To address these issues, we propose Joint Flashback Adaptation. We first introduce flashbacks -- a limited number of prompts from old tasks -- when adapting to new tasks and constrain the deviations of the model outputs compared to the original one. We then interpolate latent tasks between flashbacks and new tasks to enable jointly learning relevant latent tasks, new tasks, and flashbacks, alleviating data sparsity in flashbacks and facilitating knowledge sharing for smooth adaptation. Our method requires only a limited number of flashbacks without access to the replay data and is task-agnostic. We conduct extensive experiments on state-of-the-art large language models across 1000+ instruction-following tasks, arithmetic reasoning tasks, and general reasoning tasks. The results demonstrate the superior performance of our method in improving generalization on new tasks and reducing forgetting in old tasks. |
| title | Joint Flashback Adaptation for Forgetting-Resistant Instruction Tuning |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2505.15467 |