Metraipolis › archive › dc_968b8ddff380

Let Training Guide Selection: Online Synthetic Data Filtering via Real-Anchored Utility — arXiv:2609.29988

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.29988
License (as archived)
CC0-1.0
Archived
2026-10-08T05:26:16.871Z
SHA-256 of the archived text
f9afe46b01645108052edff4eeeaeee9ec97b62fd76ac41244d90a2132675e53
Stable URL of this page
https://metraipolis.com/documento/dc_968b8ddff380

How to cite

Let Training Guide Selection: Online Synthetic Data Filtering via Real-Anchored Utility — arXiv:2609.29988 — arXiv (en), https://arxiv.org/abs/2609.29988 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_968b8ddff380 (Metraipolis archive dc_968b8ddff380, sha256 f9afe46b01645108052edff4eeeaeee9ec97b62fd76ac41244d90a2132675e53)
BibTeX
@misc{dc968b8ddff380,
  title = {Let Training Guide Selection: Online Synthetic Data Filtering via Real-Anchored Utility — arXiv:2609.29988},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.29988},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_968b8ddff380; sha256 f9afe46b01645108052edff4eeeaeee9ec97b62fd76ac41244d90a2132675e53},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.29988. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Synthetic data can scale training supervision when real-world data are limited, but noise and distribution mismatch can reduce its value. Existing synthetic data selection methods often emphasize fidelity or diversity rather than the learner's evolving needs. We propose FROST, an online framework that estimates synthetic-data utility through gradient feedback anchored in real training data. It calibrates batch utility against recent history to determine when filtering is needed and filters samples only in out-of-band batches to determine what to retain, without an external verifier or held-out validation set. Experiments on two public benchmarks for image classification and LLM fine-tuning for text-to-SQL show that FROST filters out around 20--30% of the synthetic data while improving real-task performance compared with training on the full synthetic data pool. We further apply FROST during training in a large-scale industrial ads re-ranking system, achieving significant performance gains over a highly optimized production baseline, demonstrating its effectiveness and generalizability.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-25 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.29988 — 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 2026-10-08T04:50:39.503Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).