Metraipolis › archive › dc_0fed17e72bad

Artifact Distillation: Training-Free Transfer from Frontier to Small Language Models — arXiv:2610.37680

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.37680
License (as archived)
CC0-1.0
Archived
2026-10-31T06:10:56.026Z
SHA-256 of the archived text
9ccaad40c003c232d8b7214f4c737cdba94d163129cfa0a5c38ce366663d56d8
Stable URL of this page
https://metraipolis.com/documento/dc_0fed17e72bad

How to cite

Artifact Distillation: Training-Free Transfer from Frontier to Small Language Models — arXiv:2610.37680 — arXiv (en), https://arxiv.org/abs/2610.37680 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0fed17e72bad (Metraipolis archive dc_0fed17e72bad, sha256 9ccaad40c003c232d8b7214f4c737cdba94d163129cfa0a5c38ce366663d56d8)
BibTeX
@misc{dc0fed17e72bad,
  title = {Artifact Distillation: Training-Free Transfer from Frontier to Small Language Models — arXiv:2610.37680},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.37680},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0fed17e72bad; sha256 9ccaad40c003c232d8b7214f4c737cdba94d163129cfa0a5c38ce366663d56d8},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.37680. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Closing the gap between frontier and smaller language models is usually approached by distillation: fine-tuning the small model (student) on a frontier model's (teacher) outputs, which binds transfer to training compute and leaves the transferred knowledge largely opaque. We propose Artifact Distillation: a teacher solves training instances and externalizes its problem-solving procedure into an artifact bank that a student queries at inference time, with no gradient updates. Across four LiveBench tasks, a bank written by Claude Opus 4.5 lets Claude Haiku 4.5 outperform unaided Claude Sonnet 4.5 on spatial reasoning (0.707 vs. 0.647) at 60% of its cost, and reach Sonnet 4.5's level on table joining at 89% of its cost. The same raw bank raises Qwen3.5-27B's pooled score from 0.730 to 0.800, showing that transfer is not confined to the teacher's model family.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.37680 — 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 2026-10-31T06:07:08.577Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).