Metraipolis › archive › dc_12832cd97d8b

DRIVE: Fine-tuning via Data Contribution- and Diversity-aware Weighting with Prior Regularization — arXiv:2610.32501

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.32501
License (as archived)
CC0-1.0
Archived
2026-10-26T20:12:12.430Z
SHA-256 of the archived text
be5f7151ad1412816fa7f818e939e911c4220db7b69b908c6801b062de0d63c2
Stable URL of this page
https://metraipolis.com/documento/dc_12832cd97d8b

How to cite

DRIVE: Fine-tuning via Data Contribution- and Diversity-aware Weighting with Prior Regularization — arXiv:2610.32501 — arXiv (en), https://arxiv.org/abs/2610.32501 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_12832cd97d8b (Metraipolis archive dc_12832cd97d8b, sha256 be5f7151ad1412816fa7f818e939e911c4220db7b69b908c6801b062de0d63c2)
BibTeX
@misc{dc12832cd97d8b,
  title = {DRIVE: Fine-tuning via Data Contribution- and Diversity-aware Weighting with Prior Regularization — arXiv:2610.32501},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.32501},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_12832cd97d8b; sha256 be5f7151ad1412816fa7f818e939e911c4220db7b69b908c6801b062de0d63c2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.32501. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Single-stage fine-tuning is pervasive, yet catastrophic forgetting persists when the source pre-training data is inaccessible. Data weighting affords a natural lever to modulate sample influence. Yet existing weighting methods lack a principled mechanism to explicitly arbitrate knowledge retention against new-task adaptation, and provide no unified differentiable objective to accommodate auxiliary signals such as sample diversity and sample characteristics. This is because current schemes privilege one goal over the other, and no explicit gradient-compatible encoding exists for such signals. To address this, we propose DRIVE, a principled and efficient weight assignment framework. It derives per-sample weights via a unified differentiable objective where three coupled signals co-evolve: Shapley additivity splits contribution into anti-forgetting and adaptation components, which are reshaped by a covariance-based diversity term and stabilized by a sample-characteristic prior. Far from a mere combination, this yields an end-to-end gradient-based scheme balancing stability against plasticity. Theoretically, we prove that DRIVE enjoys superior stability-plasticity guarantees over existing weighting schemes and enhances parameter estimation robustness. Empirically, it attains superior overall performance, faster convergence, and stronger robustness. Code at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.32501 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-26T19:49:03.405Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).