Metraipolis › archive › dc_6cae673cfb1d

Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models — arXiv:2609.34933

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34933
License (as archived)
CC0-1.0
Archived
2026-10-07T21:54:17.400Z
SHA-256 of the archived text
d8197f9ab4d043f459e963cd0a853d773559b46d57430334bd2882ea96a49c87
Stable URL of this page
https://metraipolis.com/documento/dc_6cae673cfb1d

How to cite

Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models — arXiv:2609.34933 — arXiv (en), https://arxiv.org/abs/2609.34933 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6cae673cfb1d (Metraipolis archive dc_6cae673cfb1d, sha256 d8197f9ab4d043f459e963cd0a853d773559b46d57430334bd2882ea96a49c87)
BibTeX
@misc{dc6cae673cfb1d,
  title = {Don't Forget! Decomposing the Training Dynamics of Memorization in Language Models — arXiv:2609.34933},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34933},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6cae673cfb1d; sha256 d8197f9ab4d043f459e963cd0a853d773559b46d57430334bd2882ea96a49c87},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34933. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Memorization has been proposed as a mechanism to explain how language models fit the tail of their training distributions, but its training dynamics are not understood well. In this work, we take a fine-grained look at memorization by decomposing the loss trajectory of memorized sequences over training and model parameters. Across the Pythia family, we study memorization of duplicated training sequences (recitation) and rare ones (recollection). We find that memorization in both cases is characterized by sequence-level gradient alignment, though recitation suffers from misalignment with other training influences which causes forgetting, explaining the necessity for higher duplication of these examples. We further show that the lower model layers are the most involved in memorization and forgetting. Predicting memorization, our decomposition improves over a cross-entropy baseline, especially in larger models and early in training. Intervening on a small set of highly influential parameters we are able to ablate memorization in the final model. Together, these findings advance our understanding of how memorization develops during training and offer insights for predicting and intervening on it.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.CL).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.34933 — 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 2026-10-07T21:27:08.834Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).